Document Weatherreporter v0.13.0

docs/cli.md

@@ -96,7 +96,7 @@ period, prompt version, timezone, and status. Successful output has an absolute
   "command": "generate",
   "reportId": "today",
   "promptId": "weather.today_generated_text",
-  "promptVersion": "2.0.0",
+  "promptVersion": "2.1.0",
   "runId": "20260529T120000.000000000Z_today",
   "status": "succeeded",
   "timezone": "America/Chicago",
@@ -106,7 +106,10 @@ period, prompt version, timezone, and status. Successful output has an absolute
 
 When available, the summary also includes the effective `profileId`,
 `backendId`, `modelName`, `sourceWarnings`, `validationStatus`, requested
-`llmDebugPath`, and compact Distributor `notification` result. It does not
+`repairAttempts`, `llmDebugPath`, and compact Distributor `notification`
+result. `repairAttempts` is `0` when the initial output passed validation,
+positive when PromptKit made corrective generation calls, and omitted when
+validation did not complete. The summary does not
 include historical or transient artifact paths such as metadata, prompt input,
 raw generated text, render context, or notification receipts.
 
@@ -115,7 +118,8 @@ raw generated text, render context, or notification receipts.
 A run summary contains `command`, `batch`, `status`, `startedAt`, `finishedAt`,
 `total`, `succeeded`, `failed`, and a `reports` array. Each report item includes
 its identity, status, effective profile and model details when available,
-source warnings, validation status, and absolute `outputPath` after publication.
+source warnings, validation status, repair-attempt count when validation
+completed, and absolute `outputPath` after publication.
 The top-level summary may also contain a batch `notification` object and
 `error`. Batch status is `failed` if any report or the batch notification fails.
 The `total`, `succeeded`, and `failed` counters describe report items only, so
@@ -143,7 +147,8 @@ A comparison summary contains these fields in this order: `command`,
 successful `results[].reportPath` are absolute. `results` preserves the
 supplied profile order and each item contains `position`, `profileId`, optional
 `backendId`, `modelName`, `status`, optional `validationStatus`, optional
-`reportPath`, optional `llmDebugPath`, and optional safe `error`.
+`repairAttempts`, optional `reportPath`, optional `llmDebugPath`, and optional
+safe `error`. The repair-attempt semantics match the generate summary.
 
 The comparison status is `succeeded` only when every selected profile succeeds
 and the bundle is published. Individual profile failures still publish a
@@ -157,7 +162,8 @@ published Promptkit category; destination failures use `destination_<kind>`;
 and committed cleanup failures use `publication_cleanup` with a message that
 states whether a complete prior bundle, partial remnants, or no prior bundle
 remains, or that recovery state could not be inspected. It does not expose
-provider diagnostics, filesystem causes, or recovery paths. See the
+provider diagnostics, filesystem causes, or recovery paths. A provider HTTP
+failure may include its numeric status in the safe message. See the
 [comparison bundle contract](integrations/comparison-bundle.md) for durable
 artifact fields and failure invariants.
 

docs/config.md

@@ -203,16 +203,23 @@ for platform availability, security, and retention requirements.
 `profile` selects an ID; `profile_file` and `profile_dir` supply definitions.
 They are separate decisions. An explicit `profile` applies to every selected
 report. Otherwise Hourly selects `weather-light`, while Daily, Today, and
-Tomorrow select `weather-balanced` through their exact `2.0.0` prompt
+Tomorrow select `weather-balanced` through their exact `2.1.0` prompt
 definitions.
 
 Promptkit resolves a selected profile definition from a test or embedding
 consumer's explicit in-memory profile, then the configured `profile_file` or
 `profile_dir`, then Weatherreporter's embedded catalog, and finally Promptkit's
-built-in catalog. Sources provide complete definitions; fields are never
-merged. A matching malformed external profile fails rather than using the
-embedded definition. The [Promptkit integration guide](integrations/promptkit.md)
-owns the catalog and precedence details.
+maintained catalog. Weatherreporter's embedded `weather-*` definitions are small
+aliases of Promptkit's maintained base profiles, so Promptkit also resolves
+their inherited target and settings. A configured definition with the same ID
+as either a selected profile or an inherited base takes precedence. A matching
+malformed external profile fails rather than using the embedded definition. The
+[Promptkit integration guide](integrations/promptkit.md) owns the catalog and
+precedence details.
+
+An endpoint-only profile may intentionally have no backend identity. Profiles
+that require a direct API key are rejected before collection, while Promptkit
+resolves optional environment credential sources during execution.
 
 To replace the default Hourly definition with a local OpenAI-compatible
 endpoint, set `profile_file` to a copy of

docs/integrations/comparison-bundle.md

@@ -9,7 +9,7 @@ and retention belong to the [operations guide](../operations.md).
 ## Version And Layout
 
 The current and only supported manifest schema version is
-`weatherreporter.comparison.v1`. A bundle directory contains exactly these
+`weatherreporter.comparison.v2`. A bundle directory contains exactly these
 regular, non-symlinked files:
 
 ```text
@@ -51,7 +51,7 @@ this order:
 
 ```text
 position, profileId, backendId, modelName, status, validationStatus,
-reportPath, error
+repairAttempts, reportPath, error
 ```
 
 `startedAt` and `finishedAt` are nonzero UTC timestamps, and the latter is not
@@ -66,12 +66,16 @@ contiguous from one, profile IDs are distinct and nonblank, and
 `succeeded + failed == total`.
 
 A successful result has `status: "succeeded"`, `validationStatus: "passed"`,
+and a non-negative `repairAttempts` count,
 a `reportPath` exactly equal to the canonical `NN-profile-slug.md` filename for
 its position, total, and logical profile ID, and no `error`. A failed result has
 `status: "failed"`, no `reportPath`, and an `error` object with nonblank
 `category` and `message`. Its validation status is absent, `failed`, or
-`skipped`. Error messages are valid UTF-8 and no longer than 1,024 bytes.
-`backendId` and `validationStatus` are omitted when unavailable.
+`skipped`; it may also be `passed` when a WeatherReporter step after PromptKit
+validation failed. Error messages are valid UTF-8 and no longer than 1,024 bytes.
+`backendId` and `validationStatus` are omitted when unavailable. A failed
+result with any completed validation status must retain its non-negative
+`repairAttempts`; early operational failures omit both fields.
 
 Every successful Markdown file is declared by exactly one successful result.
 The directory contains no extra entries. Consumers can therefore verify the
@@ -88,7 +92,9 @@ case-variant, or duplicate fields; multiple JSON values; extra entries;
 symlinks; and future or otherwise unsupported versions. Treat a bundle that
 fails recognition as an ordinary directory, not as a compatible bundle.
 
-When replacing a recognized bundle, cancellation observed before the new
+Only v2 is recognized as a replaceable bundle; v1 is unsupported and must be
+moved or removed before a replacement at the same destination. When replacing
+a recognized bundle, cancellation observed before the new
 bundle is installed preserves the prior bundle rather than committing the
 replacement.
 

docs/integrations/promptkit.md

@@ -1,10 +1,12 @@
 # Promptkit Integration
 
-Weatherreporter uses Promptkit for all generated-text reports. The four logical prompts are `weather.daily_generated_text`, `weather.today_generated_text`, `weather.tomorrow_generated_text`, and `weather.hourly_generated_text`, each at version `2.0.0`. Their prompt assets, generated-text JSON Schemas, and Weatherreporter profile catalog are embedded by `internal/promptassets`.
+Weatherreporter uses Promptkit for all generated-text reports. The four logical prompts are `weather.daily_generated_text`, `weather.today_generated_text`, `weather.tomorrow_generated_text`, and `weather.hourly_generated_text`, each at version `2.1.0`. Their prompt assets, generated-text JSON Schemas, and Weatherreporter profile catalog are embedded by `internal/promptassets`.
 
 ## Logical Profile Catalog
 
-Prompt definitions select a stable Weatherreporter profile ID. The embedded definitions currently use Promptkit's `openrouter` backend:
+Prompt definitions select a stable Weatherreporter profile ID. Each embedded
+definition contains only its ID and one Promptkit base-profile reference; the
+effective execution settings resolve from Promptkit's maintained catalog:
 
 | Profile ID | Model | Reasoning effort | Timeout | Service tier | Default reports |
 | --- | --- | --- | --- | --- | --- |
@@ -12,22 +14,45 @@ Prompt definitions select a stable Weatherreporter profile ID. The embedded defi
 | `weather-balanced` | `~google/gemini-flash-latest` | `high` | 240 seconds | `flex` | Daily, Today, Tomorrow |
 | `weather-deep` | `~anthropic/claude-sonnet-latest` | `high` | 240 seconds | `flex` | None |
 
-The `~` prefix is part of each OpenRouter rolling-alias model ID. The embedded profiles intentionally omit endpoints, credentials, temperature, `top_p`, and output-token limits.
+The `~` prefix is part of each OpenRouter rolling-alias model ID. The embedded
+profiles intentionally omit endpoints, credentials, and execution settings;
+Promptkit owns inherited resolution and its provider-native defaults. Promptkit
+assembles its maintained catalog from independently versioned OpenRouter and
+Rakestrawhome modules selected by the Promptkit release. Weatherreporter does
+not import or register those catalog modules directly. The maintained
+`rakestrawhome-gemma-4-31b` profile is also available for ordinary and
+comparison selection and reports the `rakestrawhome` backend without
+Weatherreporter-specific configuration.
 
 ## Selection And Active Execution
 
-Before weather collection, Weatherreporter validates the report's exact generated-text report/schema/template catalog binding, prompt version and hash, output contract, and selected profile. Active profiles must resolve a nonblank backend and model identity. A nonblank `promptkit.profile` selects one profile ID for every report in the command; otherwise the prompt's declared default selects it. Promptkit resolves the selected definition in this order:
+Before weather collection, Weatherreporter validates the report's exact generated-text report/schema/template catalog binding, prompt version and hash, output contract, and selected profile. Active profiles must resolve a nonblank model; an endpoint-only profile may intentionally have no backend identity. A nonblank `promptkit.profile` selects one profile ID for every report in the command; otherwise the prompt's declared default selects it. Promptkit resolves the selected definition in this order:
 
 1. explicit in-memory profiles used by an embedding consumer or test;
 2. the configured `profile_file` or `profile_dir`;
 3. Weatherreporter's embedded fallback profiles; and
-4. Promptkit's built-in catalog.
+4. Promptkit's maintained catalog.
 
-A source falls through only when the selected ID is absent. Each source supplies a complete definition, so profile fields are not merged. A malformed matching operator definition is an error and does not fall back.
+A source falls through only when the selected ID is absent. Promptkit resolves a
+derived profile's base with the same source precedence, so a configured base
+can shadow a built-in base. A missing, cyclic, malformed, or incomplete
+selected inheritance chain is an error and does not fall back.
 
-Profiles that require a direct API key are unsupported; a profile that reports `APIKeyEnv` requires a nonblank value in that environment variable. Active results retain the selected logical profile ID and resolved backend and model. Ordinary errors, summaries, logs, and outputs exclude endpoints, credentials, rendered messages, schemas, request bodies, response bodies, and complete parameter maps.
+Profiles that require a direct API key are unsupported. Optional environment
+credential sources are Promptkit runtime concerns and are not checked by
+Weatherreporter during profile inspection. Active results retain the selected
+logical profile ID and resolved backend and model. Ordinary errors, summaries,
+logs, and outputs exclude endpoints, credentials, rendered messages, schemas,
+request bodies, response bodies, and complete parameter maps.
 
-Promptkit receives the YAML data package as an inline input and returns structured JSON that Weatherreporter validates before rendering its own Markdown template. Before accepting that JSON, Weatherreporter requires exactly one preparation callback and reconciles its prompt/profile/backend/model and rendered/input hashes with the inspected identity and completed result. The callback output contract and completed validation must use the report's expected JSON Schema mode and path. The package contains only reviewed prompt-facing warning summaries, never source transport or provenance details. Safe active provenance remains in memory. Content-rich diagnostics are opt-in through `--llm-debug-dir`; see [operations](../operations.md) for retention and permissions.
+Promptkit receives the YAML data package as an inline input and returns structured JSON that Weatherreporter validates before rendering its own Markdown template. Before accepting that JSON, Weatherreporter requires exactly one preparation callback and reconciles its prompt/profile/backend/model and rendered/input hashes with the inspected identity and completed result. The callback output contract and completed validation must use the report's expected JSON Schema mode and path. The package contains only reviewed prompt-facing warning summaries, never source transport or provenance details. Safe active provenance remains in memory. Content-rich diagnostics are opt-in through `--llm-debug-dir`; see [operations](../operations.md) for retention and permissions. Ordinary generation errors disclose only the safe Weatherreporter category and optional HTTP status; provider code, type, and message are written only to the explicit secure failure-debug artifact.
+
+Each embedded prompt permits one Promptkit-owned corrective generation after an
+eligible failed or explicitly empty result. This is not an application retry:
+Weatherreporter performs no provider retry, profile fallback, or request-level
+output-contract override. Promptkit reports cumulative usage and the actual
+number of corrective calls; repair exhaustion remains a completed validation
+failure.
 
 When capture is enabled, its preparation artifact projects a provider endpoint
 to its scheme and host and retains only reviewed execution settings. Provider

docs/internal/app-orchestration.md

@@ -7,7 +7,7 @@ is owned by the [CLI reference](../cli.md) and [operations guide](../operations.
 
 ## Single-Report Flow
 
-`GenerateDetailed` resolves the requested report and output destination before initializing an optional explicit debug writer. An explicit output file wins; otherwise the configured output directory is used, falling back to the captured working directory. Output preflight validates the final filename, permits only an absent or regular final destination, and validates the bounded same-directory temporary form without creating a missing parent. It then validates the report's generated-text catalog binding, exact Promptkit prompt, and selected profile before collecting weather data. The resolved profile, backend, and model are carried in the active result.
+`GenerateDetailed` resolves the requested report and output destination before initializing an optional explicit debug writer. An explicit output file wins; otherwise the configured output directory is used, falling back to the captured working directory. Output preflight validates the final filename, permits only an absent or regular final destination, and validates the bounded same-directory temporary form without creating a missing parent. It then validates the report's generated-text catalog binding, exact Promptkit prompt, and selected profile before collecting weather data. Profile inspection requires a model, permits an empty backend identity for endpoint-only profiles, and leaves inherited resolution and optional credential sources to Promptkit. The resolved profile, backend, model, and actual repair count (once a completed execution exists) are carried in the active result; the configured repair budget remains part of the exact prompt contract.
 
 The workflow builds facts, a module snapshot, briefing metadata, and the YAML prompt package in memory. It executes Promptkit only against the inspected prompt and profile, reconciles the preparation callback and completed result with that identity and the prepared report schema, validates the returned generated text, builds a render context, and renders Markdown. `fileutil` writes the completed Markdown through a same-directory temporary file, rechecks the final destination and context after close and immediately before the atomic rename. Only after that write succeeds does single-report notification run.
 

docs/internal/prepared-report.md

@@ -21,8 +21,9 @@ Before accepting generated JSON, the execution boundary reconciles the prepared
 report definition, inspected prompt hash and selected profile identity, the one
 preparation callback, and the completed Promptkit result. The callback and
 completion must agree on prompt, profile, backend, model, and rendered/input
-hashes; the callback output and completed validation must name the prepared
-report's JSON Schema. A mismatch produces no rendered Markdown and leaves
+hashes; the callback output must also carry the prepared report's configured
+repair budget, while the completed validation records the actual corrective
+calls used within that budget. A mismatch produces no rendered Markdown and leaves
 results with only the inspected safe identity.
 
 Single-report generation executes one prepared profile and publishes its

docs/internal/promptkit-adapter.md

@@ -2,9 +2,9 @@
 
 `internal/adapters/promptkit` maps Weatherreporter's project-owned executor contract to Promptkit. The CLI maps `promptkit` configuration to a `PromptExecutorConfig` and constructs one executor per action. Promptkit dependency types do not escape the adapter.
 
-The adapter supplies Weatherreporter's embedded prompt, schema, and fallback profile filesystems to each engine. Promptkit resolves configured operator profile sources, the embedded fallback catalog, and its built-in catalog; the adapter does not parse profile YAML, merge sources, or probe endpoints.
+The adapter supplies Weatherreporter's embedded prompt, schema, and fallback profile filesystems to each engine. Promptkit resolves configured operator profile sources, embedded profile aliases and their bases, and its maintained catalog. Promptkit privately assembles that catalog from the independently versioned OpenRouter and Rakestrawhome catalog modules selected by its release; Weatherreporter neither imports nor registers them. The adapter does not parse profile YAML, resolve inheritance, merge sources, inspect optional environment credentials, or probe endpoints.
 
-The adapter exposes exact prompt and profile validation plus prepared execution. It maps safe prompt identity, logical profile, effective backend/model, preparation, execution, validation, and optional debug values into `promptexec`. `Execute` passes the YAML package as an inline Promptkit input; it does not construct a filesystem URI or write a package file.
+The adapter exposes exact prompt and profile validation plus prepared execution. It maps safe prompt identity, logical profile, effective backend/model, preparation, execution, validation, and optional debug values into `promptexec`. An empty backend identity remains valid for an endpoint-only profile; a nonblank model is required. PromptKit's configured repair-call budget and the completed result's actual corrective-call count are retained, along with its cumulative provider usage and final candidate. Structured provider generation failures become project-owned redacted generation errors that retain only bounded details through explicit accessors. `Execute` passes the YAML package as an inline Promptkit input; it does not construct a filesystem URI or write a package file.
 
 The application uses the preparation callback to record active safe provenance in memory and optionally writes content-rich diagnostics only through an explicit debug writer. The adapter returns raw output for application validation and rendering. It does not retain application state, render Markdown, choose report definitions, or send Distributor notifications.
 

docs/operations.md

@@ -122,7 +122,9 @@ The destination is preflighted before prompt inspection and collection, then
 rechecked immediately before an atomic publish. A missing or empty directory
 is usable. A nonempty directory can be replaced only when `--replace` is given
 and it is recognized as a current Weatherreporter comparison bundle; ordinary
-directories, symlinks, and unsafe destinations are rejected. Cancellation and
+directories, symlinks, and unsafe destinations are rejected. Existing v1
+bundles are not recognized for replacement: move or remove them first.
+Cancellation and
 all failures before publication preserve an existing bundle, including a
 cancellation observed while a replacement is being prepared. If guarded
 restoration cannot complete, the error names the retained sibling bundle for
@@ -176,6 +178,12 @@ Preparation captures retain only the provider endpoint origin and reviewed
 execution settings. URL user information, paths, queries, fragments, and
 unrecognized provider parameters are omitted.
 
+Each run directory may contain `preparation.json` (v3), `execution.json` (v3),
+and, for a provider generation failure, `failure.json` (v1). The failure
+artifact retains the safe category, HTTP status, and provider code, type, and
+message for trusted debugging only. Ordinary command output never includes
+those provider details.
+
 Capture writes are confined to the requested root and fail if an unsafe
 filesystem component prevents secure artifact creation.
 

docs/policy/architecture.md

@@ -50,10 +50,14 @@ directly.
 - Prompts receive curated module packages, never unbounded raw weather payloads.
 - Every execution validates the exact prompt version and output contract before
   collection. The selected profile is configured explicitly or declared by the
-  prompt; unsupported direct-key profiles and missing reported credentials fail
-  before collection.
+  prompt; profiles requiring unsupported direct API keys fail before collection.
+  A profile may have an empty backend identity when it supplies an endpoint;
+  PromptKit resolves inherited profiles and optional credential sources when it
+  executes them.
 - Prompt and profile validation completes before weather collection. Raw output
   is validated before template rendering.
+- PromptKit may make at most the prompt contract's one corrective generation;
+  exhaustion is a validation rejection, not an application-level retry.
 - Comparison validates every explicit profile before collection, prepares one
   immutable report input, and delegates backend capacity to Promptkit rather
   than adding an application-wide execution limit.
@@ -62,6 +66,8 @@ directly.
 - Sensitive rendered prompts, schemas, input bodies, provider endpoints, and
   credentials never enter normal summaries or logs. They are written only to
   an explicit secure debug root when requested.
+- Provider-controlled diagnostics never enter ordinary outputs; they are
+  retained only in explicit secure failure-debug artifacts.
 
 ## Output, Notification, And Testing Invariants
 

docs/releases/v0.13.0.md

@@ -0,0 +1,64 @@
+# Weatherreporter v0.13.0
+
+This release upgrades Weatherreporter's generated-text integration through
+PromptKit v0.9.0 and adds bounded structured-output repair, inherited domain
+profiles, safer provider diagnostics, and repair provenance.
+
+## Summary
+
+Weatherreporter now permits one PromptKit-owned corrective generation when a
+model returns invalid or explicitly empty structured output. Successful and
+failed results report the actual correction count when validation completed,
+and explicit prompt-debug capture can retain bounded provider failure details
+without exposing them in ordinary output.
+
+The stable `weather-light`, `weather-balanced`, and `weather-deep` profiles now
+inherit their execution targets from PromptKit's maintained catalog. PromptKit
+v0.9.0 assembles that catalog from independently versioned OpenRouter and
+Rakestrawhome modules while preserving Weatherreporter's existing effective
+profile targets and configured-profile override behavior.
+
+## Compatibility
+
+The documented CLI, configuration fields, report IDs, report filenames,
+Markdown outputs, and Distributor integration remain compatible with v0.12.0.
+The four generated-text prompts advance to version `2.1.0`; an eligible failed
+validation may therefore make one additional provider call, increasing latency
+and usage for that execution.
+
+Comparison bundles advance from schema v1 to v2 so each completed validation
+can record `repairAttempts`. Weatherreporter intentionally recognizes only the
+current bundle version for guarded replacement. An existing v1 comparison
+bundle cannot be replaced with `--replace` by this release.
+
+## Upgrade
+
+No special action is required for ordinary generation, batch execution,
+configuration, or notification. Operators who want to reuse the destination
+of a v1 comparison bundle must first remove or relocate that old bundle, or
+select a new `--out-dir`.
+
+Operators should account for the possibility of one additional model call
+when output repair is needed. Existing configured standalone, inherited, and
+endpoint-only profiles retain their documented precedence.
+
+## Changes
+
+- Upgraded PromptKit from v0.5.0 through v0.9.0 and adopted its maintained
+  profile inheritance, optional credential-source behavior, structured repair,
+  generation-error contract, and externally maintained provider catalogs.
+- Converted the three Weatherreporter domain profiles into minimal aliases of
+  maintained PromptKit profiles without changing their effective model ladder.
+- Added one corrective-generation allowance to all four embedded prompts and
+  exposed actual repair counts in generate, batch, and comparison summaries.
+- Migrated the [comparison bundle contract](../integrations/comparison-bundle.md)
+  to v2 with strict repair-provenance validation and safe partial publication
+  after independent profile failures.
+- Added safe provider HTTP status to ordinary errors while confining provider
+  code, type, and message to explicit secure `failure.json` debug artifacts.
+- Strengthened offline coverage for local endpoint-only profiles, optional
+  credentials, concurrent comparison repair outcomes, and failure-debug
+  isolation.
+- Updated the [PromptKit integration guide](../integrations/promptkit.md),
+  [configuration reference](../config.md), [CLI reference](../cli.md), and
+  [operations guide](../operations.md) for the implemented contracts.

go.mod

@@ -6,9 +6,13 @@ require gopkg.in/yaml.v3 v3.0.1
 
 require (
 	gitea.maximumdirect.net/eric/distributor v0.5.0
-	gitea.maximumdirect.net/eric/promptkit v0.5.0
+	gitea.maximumdirect.net/eric/promptkit v0.9.0
 	github.com/santhosh-tekuri/jsonschema/v6 v6.0.2
 	golang.org/x/sys v0.45.0
 )
 
-require golang.org/x/text v0.14.0 // indirect
+require (
+	gitea.maximumdirect.net/eric/promptkit-backend-openrouter v1.0.0 // indirect
+	gitea.maximumdirect.net/eric/promptkit-backend-rakestrawhome v1.0.0 // indirect
+	golang.org/x/text v0.14.0 // indirect
+)

go.sum

@@ -1,7 +1,11 @@
 gitea.maximumdirect.net/eric/distributor v0.5.0 h1:+al7Bw+kMv6V35a3Sm5rUtCTQhwOn5b9x3RsclPMKJk=
 gitea.maximumdirect.net/eric/distributor v0.5.0/go.mod h1:G03FCFZPHpsUKC6SeMgTdbfNRpPQBdyTtDUj04e1Tu8=
-gitea.maximumdirect.net/eric/promptkit v0.5.0 h1:jnpazLyyNhWrB2xzwwtUkNUfktkTdkENTwuSPnKiYrc=
-gitea.maximumdirect.net/eric/promptkit v0.5.0/go.mod h1:R95NM6fbMDGDC0/UomgnSBP6ui2ns+8SZb8bESNvrDQ=
+gitea.maximumdirect.net/eric/promptkit v0.9.0 h1:IpvDRC8L6xRxQ9hpuyKOmMc5b6MeLTKYyx+h1YAjy08=
+gitea.maximumdirect.net/eric/promptkit v0.9.0/go.mod h1:oMJ/WUJImUtwJ5e+6MAGECPYAErAkOaKel0G+3T/b4E=
+gitea.maximumdirect.net/eric/promptkit-backend-openrouter v1.0.0 h1:lc062euk2qseO//D762i3JaFyulDNML3eQQX7DkYTho=
+gitea.maximumdirect.net/eric/promptkit-backend-openrouter v1.0.0/go.mod h1:AIa7kAu2mfrRQgcspe4L+DW51WqgnALQT60lqkEywJI=
+gitea.maximumdirect.net/eric/promptkit-backend-rakestrawhome v1.0.0 h1:j9YY7wsTVjzke2kHH4YAzpU0oUpM+x+nXwl1IeS+2eg=
+gitea.maximumdirect.net/eric/promptkit-backend-rakestrawhome v1.0.0/go.mod h1:4RNS+LILDg4JbS4Ts9Lwy1C92wauXJIbeQaalps4Koo=
 github.com/aws/aws-sdk-go-v2 v1.41.9 h1:/rYeyO2+HrMztAmxAq9++XJtFMqSIpSsNA0yDGALYq4=
 github.com/aws/aws-sdk-go-v2 v1.41.9/go.mod h1:+HsoOEX80qAVUitj1A2DhCNTjmb3edVyuDypb6LNEeo=
 github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.11 h1:h5+3VT69KUBK24grGuuA5saDJTj2IIjLb9au668Fo5I=

internal/adapters/promptkit/adapter.go

@@ -153,6 +153,7 @@ func outputContract(value promptkit.OutputContract) promptexec.OutputContract {
 		Format:         string(value.Format),
 		ValidationMode: string(value.ValidationMode),
 		SchemaPath:     value.SchemaPath,
+		RepairAttempts: value.RepairAttempts,
 	}
 }
 
@@ -193,6 +194,7 @@ func executionValue(value *promptkit.RunResult, captureDebug bool) *promptexec.E
 		promptexec.ValidationStatus(value.Validation.Status),
 		string(value.Validation.Mode),
 		value.Validation.SchemaPath,
+		value.Validation.RepairAttempts,
 		value.Validation.Errors,
 	)
 	rawOutput := []byte(nil)
@@ -203,6 +205,7 @@ func executionValue(value *promptkit.RunResult, captureDebug bool) *promptexec.E
 			promptexec.ValidationFailed,
 			string(value.Validation.Mode),
 			value.Validation.SchemaPath,
+			value.Validation.RepairAttempts,
 			[]string{"generated output exceeds the configured size limit"},
 		)
 	}
@@ -299,6 +302,16 @@ func classifyError(err error) error {
 	if errors.As(err, &capacityError) {
 		return promptexec.NewCapacityError(capacityError.BackendID, "prompt backend capacity is unavailable", err)
 	}
+	var generationError *promptkit.GenerationError
+	if errors.As(err, &generationError) {
+		return promptexec.NewGenerationError(
+			generationError.StatusCode(),
+			generationError.ProviderCode(),
+			generationError.ProviderType(),
+			generationError.ProviderMessage(),
+			err,
+		)
+	}
 	switch {
 	case errors.Is(err, promptkit.ErrInvalidConfig):
 		return promptexec.NewError(promptexec.InvalidConfiguration, "prompt executor configuration is invalid", err)

internal/adapters/promptkit/adapter_test.go

@@ -4,11 +4,15 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"net/http"
+	"net/http/httptest"
 	"os"
 	"path/filepath"
+	"reflect"
 	"strings"
 	"sync"
 	"testing"
+	"testing/fstest"
 	"time"
 
 	promptkit "gitea.maximumdirect.net/eric/promptkit"
@@ -20,12 +24,19 @@ type fakeClient struct {
 	mu       sync.Mutex
 	response *promptkit.GenerateResponse
 	err      error
+	outcomes []generationOutcome
+	next     int
 	calls    int
 	requests []promptkit.GenerateRequest
 	block    bool
 	started  chan struct{}
 }
 
+type generationOutcome struct {
+	response *promptkit.GenerateResponse
+	err      error
+}
+
 type recordingReader struct {
 	ref promptkit.ArtifactRef
 }
@@ -49,6 +60,11 @@ func (client *fakeClient) Generate(ctx context.Context, request promptkit.Genera
 	started := client.started
 	response := client.response
 	err := client.err
+	if client.next < len(client.outcomes) {
+		outcome := client.outcomes[client.next]
+		client.next++
+		response, err = outcome.response, outcome.err
+	}
 	client.mu.Unlock()
 	if started != nil {
 		started <- struct{}{}
@@ -102,13 +118,19 @@ func (client *fakeClient) request() promptkit.GenerateRequest {
 	return client.requests[0]
 }
 
+func (client *fakeClient) allRequests() []promptkit.GenerateRequest {
+	client.mu.Lock()
+	defer client.mu.Unlock()
+	return append([]promptkit.GenerateRequest(nil), client.requests...)
+}
+
 func TestInspectPromptAndProfile(t *testing.T) {
 	adapter := newTestAdapter(t, &fakeClient{})
-	inspection, err := adapter.InspectPrompt(context.Background(), "weather.daily_generated_text", "2.0.0")
+	inspection, err := adapter.InspectPrompt(context.Background(), "weather.daily_generated_text", "2.1.0")
 	if err != nil {
 		t.Fatalf("InspectPrompt() error = %v", err)
 	}
-	if inspection.PromptID != "weather.daily_generated_text" || inspection.PromptVersion != "2.0.0" || inspection.DefaultProfileID != "weather-balanced" {
+	if inspection.PromptID != "weather.daily_generated_text" || inspection.PromptVersion != "2.1.0" || inspection.DefaultProfileID != "weather-balanced" || inspection.Output.RepairAttempts != 1 {
 		t.Fatalf("inspection = %#v", inspection)
 	}
 	if len(inspection.Inputs) != 1 || inspection.Inputs[0].Name != "data_package" || !inspection.Inputs[0].Required || inspection.Inputs[0].ContentType != "application/yaml" {
@@ -175,15 +197,38 @@ model: file-light
 	}
 	assertProfile(t, fileAdapter, "weather-light", "", "file-light")
 
-	directory := testProfileDirectory(t, `id: weather-light
+	directory := testProfileDirectory(t, map[string]string{"profile.yml": `id: weather-light
 backend: local
 model: directory-light
-`)
+`})
 	directoryAdapter, err := New(Config{ProfileDirectory: directory, LocalEndpoint: "https://local-directory.example/v1"})
 	if err != nil {
 		t.Fatalf("New(profile directory) error = %v", err)
 	}
 	assertProfile(t, directoryAdapter, "weather-light", promptkit.BackendLocal, "directory-light")
+
+	derived := writeProfileFile(t, `id: weather-light
+base_profile: gemini-flash-latest
+`)
+	derivedAdapter, err := New(Config{ProfileFile: derived})
+	if err != nil {
+		t.Fatalf("New(derived profile) error = %v", err)
+	}
+	assertProfile(t, derivedAdapter, "weather-light", "openrouter", "~google/gemini-flash-latest")
+}
+
+func TestConfiguredBaseProfileOverridesEmbeddedProfileTarget(t *testing.T) {
+	directory := testProfileDirectory(t, map[string]string{
+		"deepseek.yml": `id: deepseek-4-flash
+backend: local
+model: shadowed-deepseek
+`,
+	})
+	adapter, err := New(Config{ProfileDirectory: directory, LocalEndpoint: "https://local-directory.example/v1"})
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	assertProfile(t, adapter, "weather-light", promptkit.BackendLocal, "shadowed-deepseek")
 }
 
 func TestMaintainedWeatherLightLocalProfileExampleInspectsOffline(t *testing.T) {
@@ -194,19 +239,69 @@ func TestMaintainedWeatherLightLocalProfileExampleInspectsOffline(t *testing.T)
 	assertProfile(t, adapter, "weather-light", "", "weather-local")
 }
 
+func TestMaintainedWeatherLightLocalProfileExampleExecutesThroughProductionClient(t *testing.T) {
+	t.Setenv("WEATHERREPORTER_TEST_MISSING_KEY", "")
+	for _, test := range []struct {
+		name             string
+		credentialSource string
+	}{
+		{name: "without credential source"},
+		{name: "with blank optional credential source", credentialSource: "\napi_key_env: WEATHERREPORTER_TEST_MISSING_KEY\n"},
+	} {
+		t.Run(test.name, func(t *testing.T) {
+			var authorization string
+			server := httptest.NewServer(http.HandlerFunc(func(writer http.ResponseWriter, request *http.Request) {
+				authorization = request.Header.Get("Authorization")
+				writer.Header().Set("Content-Type", "application/json")
+				_, _ = fmt.Fprintf(writer, `{"choices":[{"message":{"content":%q}}],"usage":{"prompt_tokens":12,"completion_tokens":8,"total_tokens":20}}`, validResponse().Content)
+			}))
+			defer server.Close()
+
+			example, err := os.ReadFile(filepath.Join("..", "..", "..", "examples", "weather-light-local-profile.yml"))
+			if err != nil {
+				t.Fatal(err)
+			}
+			profile := strings.Replace(string(example), "http://127.0.0.1:11434/v1", server.URL+"/v1", 1) + test.credentialSource
+			adapter, err := New(Config{ProfileFile: writeProfileFile(t, profile)})
+			if err != nil {
+				t.Fatalf("New() error = %v", err)
+			}
+			request := testExecuteRequest()
+			request.ProfileID = "weather-light"
+			var preparation promptexec.Preparation
+			result, err := adapter.Execute(context.Background(), request, func(value promptexec.Preparation, _ *promptexec.PreparationDebug) error {
+				preparation = value
+				return nil
+			})
+			if err != nil {
+				t.Fatalf("Execute() error = %v", err)
+			}
+			if authorization != "" {
+				t.Fatalf("Authorization header = %q, want absent", authorization)
+			}
+			if preparation.ProfileID != "weather-light" || preparation.BackendID != "" || preparation.ModelName != "weather-local" || preparation.Output.RepairAttempts != 1 {
+				t.Fatalf("preparation = %#v", preparation)
+			}
+			if result == nil || result.ProfileID != "weather-light" || result.BackendID != "" || result.ModelName != "weather-local" || result.Validation.Status != promptexec.ValidationPassed || result.Validation.RepairAttempts != 0 {
+				t.Fatalf("execution = %#v", result)
+			}
+		})
+	}
+}
+
 func TestProfileResolutionFallsThroughOnlyWhenTheConfiguredIDIsAbsent(t *testing.T) {
-	absentAdapter, err := New(Config{ProfileDirectory: testProfileDirectory(t, `id: other-profile
+	absentAdapter, err := New(Config{ProfileDirectory: testProfileDirectory(t, map[string]string{"profile.yml": `id: other-profile
 backend: openrouter
 model: other-model
-`)})
+`})})
 	if err != nil {
 		t.Fatalf("New(absent profile) error = %v", err)
 	}
 	assertProfile(t, absentAdapter, "weather-light", "openrouter", "deepseek/deepseek-v4-flash")
 
-	malformedAdapter, err := New(Config{ProfileDirectory: testProfileDirectory(t, `id: weather-light
+	malformedAdapter, err := New(Config{ProfileDirectory: testProfileDirectory(t, map[string]string{"profile.yml": `id: weather-light
 backend: openrouter
-`)})
+`})})
 	if err != nil {
 		t.Fatalf("New(malformed profile) error = %v", err)
 	}
@@ -215,6 +310,78 @@ backend: openrouter
 	}
 }
 
+func TestProfileResolutionReturnsConfiguredInheritanceFailures(t *testing.T) {
+	tests := []struct {
+		name     string
+		profile  string
+		profiles map[string]string
+	}{
+		{
+			name:    "missing base",
+			profile: "missing-base",
+			profiles: map[string]string{"missing.yml": `id: missing-base
+base_profile: unavailable
+`},
+		},
+		{
+			name:    "cyclic bases",
+			profile: "first",
+			profiles: map[string]string{
+				"first.yml": `id: first
+base_profile: second
+`,
+				"second.yml": `id: second
+base_profile: first
+`,
+			},
+		},
+		{
+			name:    "malformed base",
+			profile: "child",
+			profiles: map[string]string{
+				"child.yml": `id: child
+base_profile: malformed
+`,
+				"malformed.yml": `id: malformed
+base_profile: [not-a-profile]
+`,
+			},
+		},
+		{
+			name:    "incomplete target",
+			profile: "incomplete",
+			profiles: map[string]string{"incomplete.yml": `id: incomplete
+backend: openrouter
+`},
+		},
+	}
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			adapter, err := New(Config{ProfileDirectory: testProfileDirectory(t, test.profiles)})
+			if err != nil {
+				t.Fatalf("New() error = %v", err)
+			}
+			if _, err := adapter.InspectProfile(context.Background(), test.profile); err == nil {
+				t.Fatal("InspectProfile() error = nil, want configured inheritance error")
+			}
+		})
+	}
+}
+
+func TestRakestrawhomeBuiltInProfileInspectsOffline(t *testing.T) {
+	adapter, err := New(Config{})
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	profile, err := adapter.InspectProfile(context.Background(), "rakestrawhome-gemma-4-31b")
+	if err != nil {
+		t.Fatalf("InspectProfile() error = %v", err)
+	}
+	if profile.ProfileID != "rakestrawhome-gemma-4-31b" || profile.BackendID != "rakestrawhome" || profile.ModelName == "" {
+		t.Fatalf("profile = %#v", profile)
+	}
+}
+
 func TestProfileResolutionPreservesBuiltInAndExplicitPrecedence(t *testing.T) {
 	adapter, err := New(Config{})
 	if err != nil {
@@ -287,7 +454,7 @@ func TestExecuteEmbeddedHourlyProfileThroughPreparedPath(t *testing.T) {
 	}
 	request := promptexec.ExecuteRequest{
 		PromptID:      "weather.hourly_generated_text",
-		PromptVersion: "2.0.0",
+		PromptVersion: "2.1.0",
 		ProfileID:     "weather-light",
 		DataPackage:   []byte("report:\n  id: hourly\nbriefing: {}\n"),
 	}
@@ -400,6 +567,114 @@ func TestExecuteReturnsCompletedValidationRejection(t *testing.T) {
 	}
 }
 
+func TestExecuteMapsCorrectiveGenerationResults(t *testing.T) {
+	valid := `{"summary":"valid"}`
+	invalid := `{"summary":42}`
+	tests := []struct {
+		name        string
+		outcomes    []generationOutcome
+		wantStatus  promptexec.ValidationStatus
+		wantRepairs int
+		wantCalls   int
+		wantRaw     string
+		wantUsage   promptexec.TokenUsage
+	}{
+		{
+			name:        "first pass valid",
+			outcomes:    []generationOutcome{{response: generationResponse(valid, 2, 3, 5)}},
+			wantStatus:  promptexec.ValidationPassed,
+			wantRepairs: 0,
+			wantCalls:   1,
+			wantRaw:     valid,
+			wantUsage:   promptexec.TokenUsage{PromptTokens: 2, CompletionTokens: 3, TotalTokens: 5},
+		},
+		{
+			name:        "empty output repaired",
+			outcomes:    []generationOutcome{{response: generationResponse("", 2, 3, 5)}, {response: generationResponse(valid, 7, 11, 18)}},
+			wantStatus:  promptexec.ValidationPassed,
+			wantRepairs: 1,
+			wantCalls:   2,
+			wantRaw:     valid,
+			wantUsage:   promptexec.TokenUsage{PromptTokens: 9, CompletionTokens: 14, TotalTokens: 23},
+		},
+		{
+			name:        "invalid output repaired",
+			outcomes:    []generationOutcome{{response: generationResponse(invalid, 2, 3, 5)}, {response: generationResponse(valid, 7, 11, 18)}},
+			wantStatus:  promptexec.ValidationPassed,
+			wantRepairs: 1,
+			wantCalls:   2,
+			wantRaw:     valid,
+			wantUsage:   promptexec.TokenUsage{PromptTokens: 9, CompletionTokens: 14, TotalTokens: 23},
+		},
+		{
+			name:        "repair budget exhausted",
+			outcomes:    []generationOutcome{{response: generationResponse(invalid, 2, 3, 5)}, {response: generationResponse(invalid, 7, 11, 18)}},
+			wantStatus:  promptexec.ValidationFailed,
+			wantRepairs: 1,
+			wantCalls:   2,
+			wantRaw:     invalid,
+			wantUsage:   promptexec.TokenUsage{PromptTokens: 9, CompletionTokens: 14, TotalTokens: 23},
+		},
+	}
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			client := &fakeClient{outcomes: test.outcomes}
+			adapter := newRepairAdapter(t, client, "https://repair.example/v1")
+			var preparation promptexec.Preparation
+			result, err := adapter.Execute(context.Background(), repairExecuteRequest(), func(value promptexec.Preparation, _ *promptexec.PreparationDebug) error {
+				preparation = value
+				return nil
+			})
+			if err != nil {
+				t.Fatalf("Execute() error = %v", err)
+			}
+			if preparation.Output.RepairAttempts != 1 || result == nil || result.Validation.Status != test.wantStatus || result.Validation.RepairAttempts != test.wantRepairs || string(result.RawOutput) != test.wantRaw || result.Usage != test.wantUsage {
+				t.Fatalf("preparation/result = %#v/%#v", preparation, result)
+			}
+			requests := client.allRequests()
+			if len(requests) != test.wantCalls {
+				t.Fatalf("provider requests = %d, want %d", len(requests), test.wantCalls)
+			}
+			if test.wantCalls == 2 && !reflect.DeepEqual(requests[0].Target, requests[1].Target) {
+				t.Fatalf("corrective target = %#v, want same prepared identity as %#v", requests[1].Target, requests[0].Target)
+			}
+			if result.ProfileID != preparation.ProfileID || result.BackendID != preparation.BackendID || result.ModelName != preparation.ModelName || result.PromptID != preparation.PromptID || result.PromptVersion != preparation.PromptVersion || result.PromptHash != preparation.PromptHash {
+				t.Fatalf("prepared/result identity = %#v/%#v", preparation, result)
+			}
+		})
+	}
+}
+
+func TestExecuteMapsCorrectiveGenerationError(t *testing.T) {
+	const providerBody = `{"error":{"code":"repair-code","type":"repair-type","message":"repair-message"}}`
+	calls := 0
+	server := httptest.NewServer(http.HandlerFunc(func(writer http.ResponseWriter, _ *http.Request) {
+		calls++
+		if calls == 1 {
+			writer.Header().Set("Content-Type", "application/json")
+			_, _ = fmt.Fprintf(writer, `{"choices":[{"message":{"content":%q}}],"usage":{"prompt_tokens":2,"completion_tokens":3,"total_tokens":5}}`, `{"summary":42}`)
+			return
+		}
+		writer.Header().Set("Content-Type", "application/json")
+		writer.WriteHeader(http.StatusUnprocessableEntity)
+		_, _ = writer.Write([]byte(providerBody))
+	}))
+	defer server.Close()
+
+	adapter := newRepairAdapter(t, nil, server.URL)
+	result, err := adapter.Execute(context.Background(), repairExecuteRequest(), nil)
+	if result != nil || err == nil || calls != 2 {
+		t.Fatalf("result/error/calls = %#v/%v/%d", result, err, calls)
+	}
+	var generationError *promptexec.GenerationError
+	if !errors.As(err, &generationError) || generationError.StatusCode() != http.StatusUnprocessableEntity || generationError.ProviderCode() != "repair-code" || generationError.ProviderType() != "repair-type" || generationError.ProviderMessage() != "repair-message" {
+		t.Fatalf("generation error = %#v", err)
+	}
+	if strings.Contains(err.Error(), "repair-message") || !errors.Is(err, promptkit.ErrLLMGenerate) {
+		t.Fatalf("generation error = %v", err)
+	}
+}
+
 func TestExecuteDropsOversizedGeneratedOutput(t *testing.T) {
 	client := &fakeClient{response: &promptkit.GenerateResponse{Content: strings.Repeat("x", generatedtext.MaxGeneratedTextBytes+1)}}
 	adapter := newTestAdapter(t, client)
@@ -507,12 +782,12 @@ func TestNewValidatesConfiguration(t *testing.T) {
 	}
 }
 
-func TestLocalBackendAndMissingCredentialBehavior(t *testing.T) {
+func TestLocalBackendAndOptionalCredentialSourceBehavior(t *testing.T) {
 	t.Setenv("WEATHERREPORTER_TEST_MISSING_KEY", "")
-	profiles := testProfileDirectory(t, `id: local-profile
+	profiles := testProfileDirectory(t, map[string]string{"profile.yml": `id: local-profile
 backend: local
 model: local-model
-`)
+`})
 	adapter, err := newAdapterForTest(Config{
 		ProfileDirectory:      profiles,
 		LocalEndpoint:         "https://local.example/v1",
@@ -529,11 +804,11 @@ model: local-model
 		t.Fatalf("capacity classification = %v", got)
 	}
 
-	credentialProfiles := testProfileDirectory(t, `id: credential-profile
+	credentialProfiles := testProfileDirectory(t, map[string]string{"profile.yml": `id: credential-profile
 endpoint: https://profile.example/v1
 model: test-model
 api_key_env: WEATHERREPORTER_TEST_MISSING_KEY
-`)
+`})
 	client := &fakeClient{response: validResponse()}
 	credentialAdapter, err := newAdapterForTest(Config{ProfileDirectory: credentialProfiles}, client)
 	if err != nil {
@@ -546,8 +821,8 @@ api_key_env: WEATHERREPORTER_TEST_MISSING_KEY
 	request := testExecuteRequest()
 	request.ProfileID = "credential-profile"
 	result, err := credentialAdapter.Execute(context.Background(), request, nil)
-	if result != nil || promptexec.CategoryOf(err) != promptexec.MissingCredential || client.callCount() != 0 {
-		t.Fatalf("credential result/category/calls = %#v/%q/%d", result, promptexec.CategoryOf(err), client.callCount())
+	if err != nil || result == nil || client.callCount() != 1 {
+		t.Fatalf("credential result/error/calls = %#v/%v/%d", result, err, client.callCount())
 	}
 }
 
@@ -568,14 +843,14 @@ func assertProfile(t *testing.T, adapter *Adapter, id string, backend string, mo
 
 func newTestAdapterWithOptions(t *testing.T, client promptkit.LLMClient, options ...promptkit.Option) *Adapter {
 	t.Helper()
-	profiles := testProfileDirectory(t, `id: test-profile
+	profiles := testProfileDirectory(t, map[string]string{"profile.yml": `id: test-profile
 endpoint: https://profile.example/v1
 model: test-model
 temperature: 0.2
 max_tokens: 300
 top_p: 1
 timeout_seconds: 30
-`)
+`})
 	options = append(options, promptkit.WithLLMClient(client))
 	adapter, err := newAdapter(Config{ProfileDirectory: profiles, Timeout: time.Second}, options...)
 	if err != nil {
@@ -584,13 +859,15 @@ timeout_seconds: 30
 	return adapter
 }
 
-func testProfileDirectory(t *testing.T, profile string) string {
+func testProfileDirectory(t *testing.T, profiles map[string]string) string {
 	t.Helper()
-	profiles := t.TempDir()
-	if err := os.WriteFile(filepath.Join(profiles, "profile.yml"), []byte(profile), 0o600); err != nil {
+	directory := t.TempDir()
+	for name, profile := range profiles {
+		if err := os.WriteFile(filepath.Join(directory, name), []byte(profile), 0o600); err != nil {
 			t.Fatalf("write profile: %v", err)
 		}
-	return profiles
+	}
+	return directory
 }
 
 func writeProfileFile(t *testing.T, profile string) string {
@@ -605,7 +882,7 @@ func writeProfileFile(t *testing.T, profile string) string {
 func testExecuteRequest() promptexec.ExecuteRequest {
 	return promptexec.ExecuteRequest{
 		PromptID:      "weather.daily_generated_text",
-		PromptVersion: "2.0.0",
+		PromptVersion: "2.1.0",
 		ProfileID:     "test-profile",
 		DataPackage:   []byte("report:\n  id: daily\nbriefing: {}\n"),
 	}
@@ -618,6 +895,61 @@ func validResponse() *promptkit.GenerateResponse {
 	}
 }
 
+func generationResponse(content string, promptTokens int, completionTokens int, totalTokens int) *promptkit.GenerateResponse {
+	return &promptkit.GenerateResponse{
+		Content: content,
+		Usage: promptkit.TokenUsage{
+			PromptTokens:     promptTokens,
+			CompletionTokens: completionTokens,
+			TotalTokens:      totalTokens,
+		},
+	}
+}
+
+func newRepairAdapter(t *testing.T, client promptkit.LLMClient, endpoint string) *Adapter {
+	t.Helper()
+	profiles := testProfileDirectory(t, map[string]string{"profile.yml": "id: repair-profile\nendpoint: " + endpoint + "\nmodel: repair-model\n"})
+	options := []promptkit.Option{
+		promptkit.WithPromptFS(fstest.MapFS{
+			"repair.yml": &fstest.MapFile{Data: []byte(`id: weather.repair
+version: "1.0.0"
+default_profile: repair-profile
+inputs:
+  - name: data_package
+    required: true
+    content_type: application/yaml
+messages:
+  - role: user
+    content: "{{input \"data_package\"}}"
+output:
+  format: json
+  validation_mode: json_schema
+  schema_path: repair.schema.json
+  repair_attempts: 1
+`)}}, "."),
+		promptkit.WithSchemaFS(fstest.MapFS{
+			"repair.schema.json": &fstest.MapFile{Data: []byte(`{"type":"object","properties":{"summary":{"type":"string"}},"required":["summary"],"additionalProperties":false}`)},
+		}, "."),
+	}
+	if client != nil {
+		options = append(options, promptkit.WithLLMClient(client))
+	}
+	adapter, err := newAdapter(Config{ProfileDirectory: profiles}, options...)
+	if err != nil {
+		t.Fatalf("newAdapter() error = %v", err)
+	}
+	return adapter
+}
+
+func repairExecuteRequest() promptexec.ExecuteRequest {
+	return promptexec.ExecuteRequest{
+		PromptID:      "weather.repair",
+		PromptVersion: "1.0.0",
+		ProfileID:     "repair-profile",
+		DataPackage:   []byte("report: repair\n"),
+	}
+}
+
 func hourlyValidResponse() *promptkit.GenerateResponse {
 	return &promptkit.GenerateResponse{
 		Content: `{"summary":"A quiet hour is expected.","forecast_discussion":"Conditions remain settled.","precipitation_timing":""}`,

internal/app/app.go

@@ -89,6 +89,7 @@ type ReportResult struct {
 	ModelName        string
 	SourceWarnings   []weatherdata.SourceWarning
 	ValidationStatus promptexec.ValidationStatus
+	RepairAttempts   *int
 	LLMDebugPath     string
 	OutputPath       string
 	Notification     *NotificationResult
@@ -139,6 +140,7 @@ type BatchReportResult struct {
 	ModelName        string                      `json:"modelName,omitempty"`
 	SourceWarnings   []weatherdata.SourceWarning `json:"sourceWarnings,omitempty"`
 	ValidationStatus promptexec.ValidationStatus `json:"validationStatus,omitempty"`
+	RepairAttempts   *int                        `json:"repairAttempts,omitempty"`
 	LLMDebugPath     string                      `json:"llmDebugPath,omitempty"`
 	OutputPath       string                      `json:"outputPath,omitempty"`
 }
@@ -457,6 +459,9 @@ func copyBatchReportDetails(item *BatchReportResult, result *ReportResult) {
 	item.Timezone = result.Timezone
 	item.SourceWarnings = append([]weatherdata.SourceWarning(nil), result.SourceWarnings...)
 	item.ValidationStatus = result.ValidationStatus
+	if result.RepairAttempts != nil {
+		item.RepairAttempts = repairAttemptsPointer(*result.RepairAttempts)
+	}
 }
 
 func batchInspectionCandidates(req BatchRequest, now time.Time) ([]report.Resolved, error) {

internal/app/comparison.go

@@ -3,7 +3,6 @@ package app
 import (
 	"context"
 	"fmt"
-	"os"
 	"path/filepath"
 	"time"
 
@@ -60,6 +59,7 @@ type ComparisonProfileResult struct {
 	ModelName        string
 	Status           string
 	ValidationStatus promptexec.ValidationStatus
+	RepairAttempts   *int
 	ReportPath       string
 	LLMDebugPath     string
 	Error            *comparison.SafeError
@@ -121,7 +121,7 @@ func compareDetailed(ctx context.Context, req ComparisonRequest, publish compari
 	}
 	defer func() { _ = debugWriter.Close() }()
 	inspection, err := InspectComparisonExecution(ctx, ComparisonInspectionRequest{
-		Resolved: resolved, ProfileIDs: req.ProfileIDs, Executor: req.Executor, LookupEnv: os.LookupEnv,
+		Resolved: resolved, ProfileIDs: req.ProfileIDs, Executor: req.Executor,
 	})
 	result.PromptID, result.PromptVersion, result.PromptHash = inspection.PromptID, inspection.PromptVersion, inspection.PromptHash
 	if err != nil {
@@ -198,6 +198,9 @@ func copyComparisonOutcomes(result *ComparisonResult, outcomes []comparisonProfi
 			Position: outcome.Position, ProfileID: outcome.ProfileID, BackendID: outcome.BackendID, ModelName: outcome.ModelName,
 			Status: outcome.Status, ValidationStatus: outcome.ValidationStatus, LLMDebugPath: outcome.LLMDebugPath, Error: outcome.Error,
 		}
+		if outcome.RepairAttempts != nil {
+			profile.RepairAttempts = repairAttemptsPointer(*outcome.RepairAttempts)
+		}
 		if published && outcome.Status == comparison.StatusSucceeded {
 			profile.ReportPath = filepath.Join(result.OutputDirectory, outcome.ReportPath)
 		}
@@ -227,6 +230,9 @@ func comparisonBundle(result *ComparisonResult, dataPackage []byte, outcomes []c
 			Position: outcome.Position, ProfileID: outcome.ProfileID, BackendID: outcome.BackendID, ModelName: outcome.ModelName,
 			Status: outcome.Status, ValidationStatus: string(outcome.ValidationStatus), Error: outcome.Error,
 		}
+		if outcome.RepairAttempts != nil {
+			manifestResult.RepairAttempts = repairAttemptsPointer(*outcome.RepairAttempts)
+		}
 		if outcome.Status == comparison.StatusSucceeded {
 			manifestResult.ReportPath = outcome.ReportPath
 			bundle.Reports = append(bundle.Reports, comparison.BundleReport{Position: outcome.Position, Path: outcome.ReportPath, Markdown: outcome.Markdown})

internal/app/comparison_execution.go

@@ -31,6 +31,7 @@ type comparisonProfileOutcome struct {
 	ModelName        string
 	Status           string
 	ValidationStatus promptexec.ValidationStatus
+	RepairAttempts   *int
 	ReportPath       string
 	Markdown         []byte
 	LLMDebugPath     string
@@ -108,6 +109,9 @@ func executeComparisonProfile(ctx context.Context, req comparisonExecutionReques
 	})
 	outcome.ProfileID, outcome.BackendID, outcome.ModelName = execution.ProfileID, execution.BackendID, execution.ModelName
 	outcome.ValidationStatus = execution.ValidationStatus
+	if execution.RepairAttempts != nil {
+		outcome.RepairAttempts = repairAttemptsPointer(*execution.RepairAttempts)
+	}
 	outcome.LLMDebugPath = execution.LLMDebugPath
 	if err != nil {
 		outcome.canceled = cancellationError(err)
@@ -169,9 +173,14 @@ func comparisonExecutionMessage(err error) string {
 	if errors.Is(err, context.DeadlineExceeded) {
 		return "profile execution deadline exceeded"
 	}
+	operation := "profile execution"
 	var execution *profileExecutionError
 	if errors.As(err, &execution) {
-		return comparison.TruncateErrorMessage(execution.operation + " failed")
+		operation = execution.operation
 	}
-	return "profile execution failed"
+	var generation *promptexec.GenerationError
+	if errors.As(err, &generation) && generation.StatusCode() > 0 {
+		return comparison.TruncateErrorMessage(fmt.Sprintf("%s failed (HTTP %d)", operation, generation.StatusCode()))
+	}
+	return comparison.TruncateErrorMessage(operation + " failed")
 }

internal/app/comparison_execution_test.go

@@ -4,9 +4,11 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"net/http"
 	"os"
 	"path/filepath"
 	"reflect"
+	"strings"
 	"sync"
 	"testing"
 	"time"
@@ -72,6 +74,66 @@ func TestExecuteComparisonProfilesContinuesAfterProfileFailure(t *testing.T) {
 	}
 }
 
+func TestExecuteComparisonProfilesPreservesIndependentRepairOutcomes(t *testing.T) {
+	prepared, prompt := preparedDailyProfile(t)
+	profiles := comparisonProfiles(4)
+	executor := newBarrierExecutor(profiles)
+	executor.setValidation(profiles[0].ProfileID, promptexec.ValidationPassed, 0)
+	executor.setValidation(profiles[1].ProfileID, promptexec.ValidationPassed, 1)
+	executor.setValidation(profiles[2].ProfileID, promptexec.ValidationFailed, 1)
+	executor.setError(profiles[3].ProfileID, errors.New("provider failure"))
+	results := startComparisonExecution(t, context.Background(), comparisonExecutionRequest{
+		Prepared: prepared, Inspection: comparisonInspection(prompt, profiles), ComparisonID: "comparison_daily", Executor: executor,
+	}, executor)
+	waitForProfileStarts(t, executor, profiles, results)
+	executor.releaseAll()
+	result := <-results
+	wantStatuses := []string{comparison.StatusSucceeded, comparison.StatusSucceeded, comparison.StatusFailed, comparison.StatusFailed}
+	wantValidations := []promptexec.ValidationStatus{promptexec.ValidationPassed, promptexec.ValidationPassed, promptexec.ValidationFailed, ""}
+	wantRepairs := []*int{intPointer(0), intPointer(1), intPointer(1), nil}
+	for index, outcome := range result.Outcomes {
+		if outcome.Status != wantStatuses[index] || outcome.ValidationStatus != wantValidations[index] || !reflect.DeepEqual(outcome.RepairAttempts, wantRepairs[index]) {
+			t.Fatalf("outcome[%d] = %#v, want status/validation/repairs %q/%q/%#v", index, outcome, wantStatuses[index], wantValidations[index], wantRepairs[index])
+		}
+	}
+}
+
+func TestExecuteComparisonProfilesCapturesConcurrentProviderFailures(t *testing.T) {
+	prepared, prompt := preparedDailyProfile(t)
+	profiles := comparisonProfiles(2)
+	debugWriter, err := promptdebug.NewPromptDebugWriter(t.TempDir())
+	if errors.Is(err, promptdebug.ErrSecureCaptureUnsupported) {
+		t.Skipf("secure prompt debug capture is unavailable: %v", err)
+	}
+	if err != nil {
+		t.Fatalf("NewPromptDebugWriter() error = %v", err)
+	}
+	markers := []string{"first-provider-private-marker", "second-provider-private-marker"}
+	statuses := []int{http.StatusTooManyRequests, http.StatusServiceUnavailable}
+	executor := newBarrierExecutor(profiles)
+	for index, profile := range profiles {
+		executor.setError(profile.ProfileID, promptexec.NewGenerationError(statuses[index], "provider_code", "provider_type", markers[index], nil))
+	}
+	results := startComparisonExecution(t, context.Background(), comparisonExecutionRequest{
+		Prepared: prepared, Inspection: comparisonInspection(prompt, profiles), ComparisonID: "comparison_daily", DebugWriter: debugWriter, Executor: executor,
+	}, executor)
+	waitForProfileStarts(t, executor, profiles, results)
+	executor.releaseAll()
+	result := <-results
+	for index, outcome := range result.Outcomes {
+		if outcome.Status != comparison.StatusFailed || outcome.Error == nil || outcome.Error.Category != string(promptexec.Generation) || outcome.Error.Message != fmt.Sprintf("execute prompt failed (HTTP %d)", statuses[index]) || strings.Contains(outcome.Error.Message, markers[index]) || outcome.LLMDebugPath == "" {
+			t.Fatalf("outcome[%d] = %#v", index, outcome)
+		}
+		failure, readErr := os.ReadFile(filepath.Join(outcome.LLMDebugPath, "failure.json"))
+		if readErr != nil {
+			t.Fatal(readErr)
+		}
+		if !strings.Contains(string(failure), markers[index]) || strings.Contains(string(failure), markers[1-index]) {
+			t.Fatalf("failure[%d] = %s", index, failure)
+		}
+	}
+}
+
 func TestExecuteComparisonProfilesPropagatesCancellationAndJoins(t *testing.T) {
 	prepared, prompt := preparedDailyProfile(t)
 	profiles := comparisonProfiles(4)
@@ -139,6 +201,8 @@ type barrierExecutor struct {
 	releases         map[string]chan struct{}
 	requests         map[string]promptexec.ExecuteRequest
 	errors           map[string]error
+	validations      map[string]promptexec.ValidationStatus
+	repairAttempts   map[string]int
 	profiles         map[string]ComparisonProfileInspection
 	inFlight         int
 	maximum          int
@@ -153,7 +217,7 @@ func newBarrierExecutor(profiles []ComparisonProfileInspection) *barrierExecutor
 	}
 	return &barrierExecutor{
 		started: make(chan string, len(profiles)), callbackFailures: make(chan error, len(profiles)), releases: releases,
-		requests: make(map[string]promptexec.ExecuteRequest, len(profiles)), errors: map[string]error{}, profiles: identities,
+		requests: make(map[string]promptexec.ExecuteRequest, len(profiles)), errors: map[string]error{}, validations: map[string]promptexec.ValidationStatus{}, repairAttempts: map[string]int{}, profiles: identities,
 	}
 }
 
@@ -170,7 +234,8 @@ func (e *barrierExecutor) Execute(ctx context.Context, req promptexec.ExecuteReq
 	e.mu.Lock()
 	profile := e.profiles[req.ProfileID]
 	e.mu.Unlock()
-	if err := callback(promptexec.Preparation{PromptID: req.PromptID, PromptVersion: req.PromptVersion, PromptHash: generationPromptHash, ProfileID: req.ProfileID, BackendID: profile.BackendID, ModelName: profile.ModelName, Output: promptexec.OutputContract{Format: "json", ValidationMode: "json_schema", SchemaPath: generationDefinitionForPrompt(req.PromptID).GeneratedTextSchemaID + ".generated_text.schema.json"}, StartedAt: stamp, EndedAt: stamp}, nil); err != nil {
+	definition := generationDefinitionForPrompt(req.PromptID)
+	if err := callback(promptexec.Preparation{PromptID: req.PromptID, PromptVersion: req.PromptVersion, PromptHash: generationPromptHash, ProfileID: req.ProfileID, BackendID: profile.BackendID, ModelName: profile.ModelName, Output: promptexec.OutputContract{Format: "json", ValidationMode: "json_schema", SchemaPath: definition.GeneratedTextSchemaID + ".generated_text.schema.json", RepairAttempts: definition.GeneratedTextRepairAttempts}, StartedAt: stamp, EndedAt: stamp}, nil); err != nil {
 		e.callbackFailures <- err
 		return nil, err
 	}
@@ -194,15 +259,20 @@ func (e *barrierExecutor) Execute(ctx context.Context, req promptexec.ExecuteReq
 	e.mu.Lock()
 	e.inFlight--
 	err := e.errors[req.ProfileID]
+	validationStatus := e.validations[req.ProfileID]
+	repairAttempts := e.repairAttempts[req.ProfileID]
 	e.mu.Unlock()
 	if err != nil {
 		return nil, err
 	}
+	if validationStatus == "" {
+		validationStatus = promptexec.ValidationPassed
+	}
 	return &promptexec.Execution{
 		PromptID: req.PromptID, PromptVersion: req.PromptVersion, PromptHash: generationPromptHash,
 		ProfileID: req.ProfileID, BackendID: profile.BackendID, ModelName: profile.ModelName,
 		StartedAt: stamp, EndedAt: stamp, RawOutput: comparisonRawOutput(),
-		Validation: promptexec.NewValidation(promptexec.ValidationPassed, "json_schema", generationDefinitionForPrompt(req.PromptID).GeneratedTextSchemaID+".generated_text.schema.json", nil),
+		Validation: promptexec.NewValidation(validationStatus, "json_schema", generationDefinitionForPrompt(req.PromptID).GeneratedTextSchemaID+".generated_text.schema.json", repairAttempts, nil),
 	}, nil
 }
 
@@ -219,6 +289,13 @@ func (e *barrierExecutor) setError(profileID string, err error) {
 	e.errors[profileID] = err
 }
 
+func (e *barrierExecutor) setValidation(profileID string, status promptexec.ValidationStatus, repairAttempts int) {
+	e.mu.Lock()
+	defer e.mu.Unlock()
+	e.validations[profileID] = status
+	e.repairAttempts[profileID] = repairAttempts
+}
+
 func (e *barrierExecutor) release(profileID string) {
 	close(e.releases[profileID])
 }
@@ -317,4 +394,8 @@ func bytesEqual(left, right []byte) bool {
 	return reflect.DeepEqual(left, right)
 }
 
+func intPointer(value int) *int {
+	return &value
+}
+
 var _ promptexec.Executor = (*barrierExecutor)(nil)

internal/app/comparison_test.go

@@ -87,6 +87,40 @@ func TestCompareDetailedPublishesPartialBundleAndReturnsAggregateError(t *testin
 	}
 }
 
+func TestCompareDetailedPublishesPostValidationProfileFailure(t *testing.T) {
+	bundle := generationBundle(t)
+	executor := &generationExecutor{complete: func(execution *promptexec.Execution) {
+		if execution.ProfileID == "weather-deep" {
+			execution.RawOutput = []byte(`{"summary":42}`)
+		}
+	}}
+	result, err := CompareDetailed(context.Background(), ComparisonRequest{
+		Config: comparisonConfig(), Report: ReportDaily, ProfileIDs: []string{"weather-light", "weather-deep"},
+		WorkingDir: t.TempDir(), Date: generationTime("2026-05-29T12:00:00-05:00"),
+		Clock:     timeutil.FixedClock{Time: generationTime("2026-05-29T08:30:00-05:00")},
+		Collector: &generationCollector{bundle: &bundle}, Executor: executor,
+	})
+	if err == nil || result == nil || result.Succeeded != 1 || result.Failed != 1 || result.ManifestPath == "" {
+		t.Fatalf("CompareDetailed() result/error = %#v/%v", result, err)
+	}
+	failure := result.Results[1]
+	if failure.Status != comparison.StatusFailed || failure.ValidationStatus != promptexec.ValidationPassed || failure.RepairAttempts == nil || *failure.RepairAttempts != 0 {
+		t.Fatalf("post-validation failure = %#v", failure)
+	}
+	data, readErr := os.ReadFile(result.ManifestPath)
+	if readErr != nil {
+		t.Fatal(readErr)
+	}
+	var manifest comparison.Manifest
+	if decodeErr := json.Unmarshal(data, &manifest); decodeErr != nil {
+		t.Fatal(decodeErr)
+	}
+	manifestFailure := manifest.Results[1]
+	if manifestFailure.ValidationStatus != "passed" || manifestFailure.RepairAttempts == nil || *manifestFailure.RepairAttempts != 0 {
+		t.Fatalf("published post-validation failure = %#v", manifestFailure)
+	}
+}
+
 func TestCompareDetailedRetainsCommittedPathsWhenBackupCleanupFails(t *testing.T) {
 	for _, test := range []struct {
 		name  string

internal/app/generation_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"encoding/json"
 	"errors"
+	"net/http"
 	"os"
 	"path/filepath"
 	"strings"
@@ -68,6 +69,7 @@ type generationExecutor struct {
 	beforeExecute        func(promptexec.ExecuteRequest)
 	cancelBeforeReturn   context.CancelFunc
 	validation           promptexec.ValidationStatus
+	repairAttempts       int
 	validations          map[string]promptexec.ValidationStatus
 	rawOutput            []byte
 	waitForCancellation  map[string]bool
@@ -88,7 +90,7 @@ func (e *generationExecutor) InspectPrompt(_ context.Context, id, version string
 		return promptexec.PromptInspection{}, e.inspectErr
 	}
 	definition := generationDefinitionForPrompt(id)
-	return promptexec.PromptInspection{PromptID: id, PromptVersion: version, PromptHash: generationPromptHash, DefaultProfileID: "fixture", Inputs: []promptexec.InputDefinition{{Name: "data_package", Required: true, ContentType: "application/yaml"}}, Output: promptexec.OutputContract{Format: "json", ValidationMode: "json_schema", SchemaPath: definition.GeneratedTextSchemaID + ".generated_text.schema.json"}}, nil
+	return promptexec.PromptInspection{PromptID: id, PromptVersion: version, PromptHash: generationPromptHash, DefaultProfileID: "fixture", Inputs: []promptexec.InputDefinition{{Name: "data_package", Required: true, ContentType: "application/yaml"}}, Output: promptexec.OutputContract{Format: "json", ValidationMode: "json_schema", SchemaPath: definition.GeneratedTextSchemaID + ".generated_text.schema.json", RepairAttempts: definition.GeneratedTextRepairAttempts}}, nil
 }
 func (e *generationExecutor) InspectProfile(_ context.Context, id string) (promptexec.ProfileInspection, error) {
 	generationExecutorMu.Lock()
@@ -112,7 +114,8 @@ func (e *generationExecutor) Execute(ctx context.Context, req promptexec.Execute
 			calls = 1
 		}
 		for range calls {
-			preparation := promptexec.Preparation{PromptID: req.PromptID, PromptVersion: req.PromptVersion, PromptHash: generationPromptHash, RenderedPromptHash: "rendered-hash", ProfileID: req.ProfileID, BackendID: "fixture", ModelName: "fixture-model", Output: promptexec.OutputContract{Format: "json", ValidationMode: "json_schema", SchemaPath: generationDefinitionForPrompt(req.PromptID).GeneratedTextSchemaID + ".generated_text.schema.json"}, StartedAt: stamp, EndedAt: stamp}
+			definition := generationDefinitionForPrompt(req.PromptID)
+			preparation := promptexec.Preparation{PromptID: req.PromptID, PromptVersion: req.PromptVersion, PromptHash: generationPromptHash, RenderedPromptHash: "rendered-hash", ProfileID: req.ProfileID, BackendID: "fixture", ModelName: "fixture-model", Output: promptexec.OutputContract{Format: "json", ValidationMode: "json_schema", SchemaPath: definition.GeneratedTextSchemaID + ".generated_text.schema.json", RepairAttempts: definition.GeneratedTextRepairAttempts}, StartedAt: stamp, EndedAt: stamp}
 			if prepare != nil {
 				prepare(&preparation)
 			}
@@ -128,6 +131,7 @@ func (e *generationExecutor) Execute(ctx context.Context, req promptexec.Execute
 	profileErr := e.executeErrors[req.ProfileID]
 	executeErr := e.executeErr
 	status := e.validation
+	repairAttempts := e.repairAttempts
 	if profileStatus, ok := e.validations[req.ProfileID]; ok {
 		status = profileStatus
 	}
@@ -162,7 +166,7 @@ func (e *generationExecutor) Execute(ctx context.Context, req promptexec.Execute
 	if cancelBeforeReturn != nil {
 		cancelBeforeReturn()
 	}
-	execution := &promptexec.Execution{RunID: "provider-run", PromptID: req.PromptID, PromptVersion: req.PromptVersion, PromptHash: generationPromptHash, RenderedPromptHash: "rendered-hash", ProfileID: req.ProfileID, BackendID: "fixture", ModelName: "fixture-model", StartedAt: stamp, EndedAt: stamp, RawOutput: rawOutput, Validation: promptexec.NewValidation(status, "json_schema", generationDefinitionForPrompt(req.PromptID).GeneratedTextSchemaID+".generated_text.schema.json", nil)}
+	execution := &promptexec.Execution{RunID: "provider-run", PromptID: req.PromptID, PromptVersion: req.PromptVersion, PromptHash: generationPromptHash, RenderedPromptHash: "rendered-hash", ProfileID: req.ProfileID, BackendID: "fixture", ModelName: "fixture-model", StartedAt: stamp, EndedAt: stamp, RawOutput: rawOutput, Validation: promptexec.NewValidation(status, "json_schema", generationDefinitionForPrompt(req.PromptID).GeneratedTextSchemaID+".generated_text.schema.json", repairAttempts, nil)}
 	if complete != nil {
 		complete(execution)
 	}
@@ -546,6 +550,66 @@ func TestGenerateDetailedWritesRequestedPromptDebugArtifacts(t *testing.T) {
 	}
 }
 
+func TestGenerateDetailedCapturesProviderFailureOnlyInDebugArtifacts(t *testing.T) {
+	bundle := generationBundle(t)
+	debugRoot := t.TempDir()
+	const marker = "provider-private-generation-marker"
+	result, err := GenerateDetailed(context.Background(), GenerateRequest{
+		Config: generationConfig(), Report: ReportDaily,
+		Date: generationTime("2026-05-29T12:00:00-05:00"), Now: generationTime("2026-05-29T08:30:00-05:00"),
+		WorkingDir: t.TempDir(), LLMDebugDir: debugRoot, Collector: &generationCollector{bundle: &bundle},
+		Executor: &generationExecutor{executeErr: promptexec.NewGenerationError(http.StatusTooManyRequests, "rate_limit", "provider_error", marker, nil)},
+	})
+	if errors.Is(err, promptdebug.ErrSecureCaptureUnsupported) {
+		t.Skipf("secure prompt debug capture is unavailable: %v", err)
+	}
+	if err == nil || result == nil || result.LLMDebugPath == "" || promptexec.CategoryOf(err) != promptexec.Generation || !strings.Contains(err.Error(), "HTTP 429") || strings.Contains(err.Error(), marker) || result.OutputPath != "" {
+		t.Fatalf("GenerateDetailed() result/error = %#v/%v", result, err)
+	}
+	for _, name := range []string{"preparation.json", "failure.json"} {
+		if _, statErr := os.Stat(filepath.Join(result.LLMDebugPath, name)); statErr != nil {
+			t.Fatalf("debug artifact %q: %v", name, statErr)
+		}
+	}
+	data, readErr := os.ReadFile(filepath.Join(result.LLMDebugPath, "failure.json"))
+	if readErr != nil || !strings.Contains(string(data), marker) {
+		t.Fatalf("failure artifact = %q, error = %v", data, readErr)
+	}
+}
+
+func TestGenerateDetailedPreservesProviderFailureWhenFailureDebugWriteFails(t *testing.T) {
+	bundle := generationBundle(t)
+	debugRoot := t.TempDir()
+	const marker = "provider-private-write-failure-marker"
+	var setupErr error
+	executor := &generationExecutor{
+		executeErr: promptexec.NewGenerationError(http.StatusServiceUnavailable, "unavailable", "provider_error", marker, nil),
+		beforeExecute: func(promptexec.ExecuteRequest) {
+			setupErr = filepath.Walk(debugRoot, func(path string, info os.FileInfo, err error) error {
+				if err != nil {
+					return err
+				}
+				if info.Name() == "preparation.json" {
+					return os.Mkdir(filepath.Join(filepath.Dir(path), "failure.json"), 0o700)
+				}
+				return nil
+			})
+		},
+	}
+	result, err := GenerateDetailed(context.Background(), GenerateRequest{
+		Config: generationConfig(), Report: ReportDaily,
+		Date: generationTime("2026-05-29T12:00:00-05:00"), Now: generationTime("2026-05-29T08:30:00-05:00"),
+		WorkingDir: t.TempDir(), LLMDebugDir: debugRoot, Collector: &generationCollector{bundle: &bundle}, Executor: executor,
+	})
+	if errors.Is(err, promptdebug.ErrSecureCaptureUnsupported) {
+		t.Skipf("secure prompt debug capture is unavailable: %v", err)
+	}
+	var generationError *promptexec.GenerationError
+	if setupErr != nil || err == nil || result == nil || result.OutputPath != "" || promptexec.CategoryOf(err) != promptexec.Generation || !errors.As(err, &generationError) || generationError.StatusCode() != http.StatusServiceUnavailable || !strings.Contains(err.Error(), "HTTP 503") || strings.Contains(err.Error(), marker) {
+		t.Fatalf("GenerateDetailed() setup/result/error = %v/%#v/%v", setupErr, result, err)
+	}
+}
+
 func generationConfig() config.Config {
 	cfg := config.Defaults()
 	cfg.WeatherAPI.Timezone, cfg.Location.ID = "America/Chicago", "home"

internal/app/profile_execution.go

@@ -2,6 +2,7 @@ package app
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"reflect"
 
@@ -24,6 +25,7 @@ type profileExecutionOutcome struct {
 	BackendID        string
 	ModelName        string
 	ValidationStatus promptexec.ValidationStatus
+	RepairAttempts   *int
 	LLMDebugPath     string
 }
 
@@ -94,11 +96,24 @@ func executePreparedProfile(ctx context.Context, req profileExecutionRequest) (p
 		if callbackFailed {
 			return outcome, nil, &profileExecutionError{operation: "execute prompt", err: err, callbackFailure: true}
 		}
+		if req.DebugWriter != nil && req.DebugWriter.Enabled() && req.DebugRef != nil {
+			var generationError *promptexec.GenerationError
+			if errors.As(err, &generationError) {
+				path, debugErr := req.DebugWriter.WriteFailure(*req.DebugRef, generationError)
+				if path != "" {
+					outcome.LLMDebugPath = path
+				}
+				if debugErr != nil {
+					err = errors.Join(err, promptDebugWriteError(debugErr))
+				}
+			}
+		}
 		return outcome, nil, &profileExecutionError{operation: "execute prompt", err: classifiedPromptError("prompt execution failed", err)}
 	}
 	if execution == nil {
 		return outcome, nil, &profileExecutionError{operation: "execute prompt", err: promptexec.NewError(promptexec.Generation, "prompt executor returned no execution", nil)}
 	}
+	outcome.RepairAttempts = repairAttemptsPointer(execution.Validation.RepairAttempts)
 	if preparationCount != 1 {
 		return outcome, nil, &profileExecutionError{operation: "validate prompt provenance", err: promptProvenanceError()}
 	}
@@ -157,7 +172,7 @@ func validatePreparedExecutionRequest(req profileExecutionRequest) error {
 	if req.Prompt.ProfileID != "" && (req.Prompt.ProfileID != req.Profile.ProfileID || req.Prompt.BackendID != req.Profile.BackendID || req.Prompt.ModelName != req.Profile.ModelName) {
 		return promptProvenanceError()
 	}
-	if req.Prompt.PromptHash == "" || req.Profile.ProfileID == "" || req.Profile.BackendID == "" || req.Profile.ModelName == "" {
+	if req.Prompt.PromptHash == "" || req.Profile.ProfileID == "" || req.Profile.ModelName == "" {
 		return promptProvenanceError()
 	}
 	return nil
@@ -178,12 +193,19 @@ func validateExecutionProvenance(req profileExecutionRequest, preparation prompt
 	if execution.PromptID != preparation.PromptID || execution.PromptVersion != preparation.PromptVersion || execution.PromptHash != preparation.PromptHash ||
 		execution.RenderedPromptHash != preparation.RenderedPromptHash || !reflect.DeepEqual(execution.InputHashes, preparation.InputHashes) ||
 		execution.ProfileID != preparation.ProfileID || execution.BackendID != preparation.BackendID || execution.ModelName != preparation.ModelName ||
-		execution.Validation.Mode != "json_schema" || execution.Validation.SchemaPath != definition.GeneratedTextSchemaID+".generated_text.schema.json" {
+		execution.Validation.Mode != "json_schema" || execution.Validation.SchemaPath != definition.GeneratedTextSchemaID+".generated_text.schema.json" ||
+		execution.Validation.RepairAttempts < 0 || execution.Validation.RepairAttempts > preparation.Output.RepairAttempts ||
+		preparation.Output.RepairAttempts != definition.GeneratedTextRepairAttempts {
 		return promptProvenanceError()
 	}
 	return nil
 }
 
+func repairAttemptsPointer(value int) *int {
+	copy := value
+	return &copy
+}
+
 func promptProvenanceError() error {
 	return promptexec.NewError(promptexec.InvalidConfiguration, "prompt execution provenance is inconsistent", nil)
 }

internal/app/profile_execution_test.go

@@ -26,7 +26,7 @@ func TestExecutePreparedProfileRendersWithoutPublishing(t *testing.T) {
 	if err != nil {
 		t.Fatalf("executePreparedProfile() error = %v", err)
 	}
-	if len(rendered) == 0 || outcome.ValidationStatus != promptexec.ValidationPassed || outcome.ProfileID != inspection.ProfileID || executor.executeCalls != 1 {
+	if len(rendered) == 0 || outcome.ValidationStatus != promptexec.ValidationPassed || outcome.RepairAttempts == nil || *outcome.RepairAttempts != 0 || outcome.ProfileID != inspection.ProfileID || executor.executeCalls != 1 {
 		t.Fatalf("outcome/rendered/execution calls = %#v/%q/%d", outcome, rendered, executor.executeCalls)
 	}
 	if _, statErr := os.Stat(outputPath); !os.IsNotExist(statErr) {
@@ -34,6 +34,20 @@ func TestExecutePreparedProfileRendersWithoutPublishing(t *testing.T) {
 	}
 }
 
+func TestExecutePreparedProfileRetainsCompletedRepairAttemptsOnLaterFailure(t *testing.T) {
+	prepared, inspection := preparedDailyProfile(t)
+	prepared.resolved.Definition.GeneratedTextRepairAttempts = 1
+	executor := &generationExecutor{repairAttempts: 1, rawOutput: []byte(`{"summary":42}`), prepare: func(value *promptexec.Preparation) { value.Output.RepairAttempts = 1 }}
+	outcome, _, err := executePreparedProfile(context.Background(), profileExecutionRequest{
+		Prepared: prepared, Prompt: inspection,
+		Profile:  promptexec.ProfileInspection{ProfileID: inspection.ProfileID, BackendID: inspection.BackendID, ModelName: inspection.ModelName},
+		Executor: executor,
+	})
+	if err == nil || outcome.RepairAttempts == nil || *outcome.RepairAttempts != 1 {
+		t.Fatalf("outcome/error = %#v/%v", outcome, err)
+	}
+}
+
 func TestExecutePreparedProfileKeepsDebugCallbackFailureLocal(t *testing.T) {
 	prepared, inspection := preparedDailyProfile(t)
 	debugWriter, err := promptdebug.NewPromptDebugWriter(t.TempDir())

internal/app/prompt_generate.go

@@ -48,6 +48,9 @@ func generatePromptReport(ctx context.Context, req promptReportRequest) (*Report
 	})
 	result.ProfileID, result.BackendID, result.ModelName = outcome.ProfileID, outcome.BackendID, outcome.ModelName
 	result.ValidationStatus = outcome.ValidationStatus
+	if outcome.RepairAttempts != nil {
+		result.RepairAttempts = repairAttemptsPointer(*outcome.RepairAttempts)
+	}
 	result.LLMDebugPath = outcome.LLMDebugPath
 	if err != nil {
 		return result, generatedProfileExecutionError(req.Resolved, result.RunID, err)

internal/app/prompt_inspection.go

@@ -2,7 +2,6 @@ package app
 
 import (
 	"context"
-	"os"
 	"strings"
 
 	"gitea.maximumdirect.net/eric/weatherreporter/internal/comparison"
@@ -18,7 +17,6 @@ type PromptInspectionRequest struct {
 	Resolved  report.Resolved
 	Executor  promptexec.Executor
 	Promptkit config.PromptkitConfig
-	LookupEnv func(string) (string, bool)
 }
 
 // PromptInspectionResult contains only safe identity and provenance from a
@@ -39,7 +37,6 @@ type PromptExecutionsInspectionRequest struct {
 	Resolved  []report.Resolved
 	Executor  promptexec.Executor
 	Promptkit config.PromptkitConfig
-	LookupEnv func(string) (string, bool)
 }
 
 // ComparisonInspectionRequest contains the explicit profile selection for one
@@ -48,7 +45,6 @@ type ComparisonInspectionRequest struct {
 	Resolved   report.Resolved
 	ProfileIDs []string
 	Executor   promptexec.Executor
-	LookupEnv  func(string) (string, bool)
 }
 
 // ComparisonInspectionResult contains the safe, shared prompt identity and
@@ -76,7 +72,6 @@ func InspectPromptExecution(ctx context.Context, req PromptInspectionRequest) (P
 		Resolved:  []report.Resolved{req.Resolved},
 		Executor:  req.Executor,
 		Promptkit: req.Promptkit,
-		LookupEnv: req.LookupEnv,
 	})
 	if err != nil {
 		return PromptInspectionResult{}, err
@@ -111,7 +106,7 @@ func InspectPromptExecutions(ctx context.Context, req PromptExecutionsInspection
 		}
 		profile, ok := profiles[profileID]
 		if !ok {
-			profile, err = inspectPromptProfile(ctx, req.Executor, profileID, req.LookupEnv)
+			profile, err = inspectPromptProfile(ctx, req.Executor, profileID)
 			if err != nil {
 				return nil, err
 			}
@@ -154,7 +149,7 @@ func InspectComparisonExecution(ctx context.Context, req ComparisonInspectionReq
 		handler:       handler,
 	}
 	for _, profileID := range req.ProfileIDs {
-		profile, err := inspectPromptProfile(ctx, req.Executor, profileID, req.LookupEnv)
+		profile, err := inspectPromptProfile(ctx, req.Executor, profileID)
 		if err != nil {
 			return result, comparisonInspectionError("comparison profile inspection failed", err)
 		}
@@ -190,7 +185,7 @@ func inspectPromptContract(ctx context.Context, executor promptexec.Executor, de
 	return inspection, nil
 }
 
-func inspectPromptProfile(ctx context.Context, executor promptexec.Executor, profileID string, lookupEnv func(string) (string, bool)) (promptexec.ProfileInspection, error) {
+func inspectPromptProfile(ctx context.Context, executor promptexec.Executor, profileID string) (promptexec.ProfileInspection, error) {
 	profile, err := executor.InspectProfile(ctx, profileID)
 	if err != nil {
 		return promptexec.ProfileInspection{}, promptInspectionError("profile inspection failed", err)
@@ -201,16 +196,7 @@ func inspectPromptProfile(ctx context.Context, executor promptexec.Executor, pro
 	if profile.CredentialRequired {
 		return promptexec.ProfileInspection{}, promptexec.NewError(promptexec.MissingCredential, "selected profile requires an unsupported direct API key", nil)
 	}
-	if strings.TrimSpace(profile.APIKeyEnv) != "" {
-		if lookupEnv == nil {
-			lookupEnv = os.LookupEnv
-		}
-		value, present := lookupEnv(profile.APIKeyEnv)
-		if !present || strings.TrimSpace(value) == "" {
-			return promptexec.ProfileInspection{}, promptexec.NewError(promptexec.MissingCredential, "selected profile credential is unavailable", nil)
-		}
-	}
-	if strings.TrimSpace(profile.BackendID) == "" || strings.TrimSpace(profile.ModelName) == "" {
+	if strings.TrimSpace(profile.ModelName) == "" {
 		return promptexec.ProfileInspection{}, promptexec.NewError(promptexec.InvalidConfiguration, "profile inspection did not return a complete execution identity", nil)
 	}
 	return profile, nil
@@ -221,7 +207,7 @@ func validPromptInput(inputs []promptexec.InputDefinition) bool {
 }
 
 func validPromptOutput(definition report.Definition, output promptexec.OutputContract) bool {
-	return output.Format == "json" && output.ValidationMode == "json_schema" && output.SchemaPath == definition.GeneratedTextSchemaID+".generated_text.schema.json"
+	return output.Format == "json" && output.ValidationMode == "json_schema" && output.SchemaPath == definition.GeneratedTextSchemaID+".generated_text.schema.json" && output.RepairAttempts == definition.GeneratedTextRepairAttempts
 }
 
 func promptInspectionError(operation string, err error) error {

internal/app/prompt_inspection_test.go

@@ -50,7 +50,6 @@ func TestInspectPromptExecutionRejectsInvalidContractsAndCredentials(t *testing.
 		name         string
 		prompt       promptexec.PromptInspection
 		profile      promptexec.ProfileInspection
-		lookupEnv    func(string) (string, bool)
 		wantCategory promptexec.ErrorCategory
 	}{
 		{
@@ -81,9 +80,9 @@ func TestInspectPromptExecutionRejectsInvalidContractsAndCredentials(t *testing.
 			wantCategory: promptexec.InvalidConfiguration,
 		},
 		{
-			name:         "missing profile backend",
+			name:         "missing profile model",
 			prompt:       basePrompt,
-			profile:      promptexec.ProfileInspection{ProfileID: "default-profile", ModelName: "model"},
+			profile:      promptexec.ProfileInspection{ProfileID: "default-profile", BackendID: "backend"},
 			wantCategory: promptexec.InvalidConfiguration,
 		},
 		{
@@ -92,18 +91,11 @@ func TestInspectPromptExecutionRejectsInvalidContractsAndCredentials(t *testing.
 			profile:      promptexec.ProfileInspection{ProfileID: "default-profile", CredentialRequired: true},
 			wantCategory: promptexec.MissingCredential,
 		},
-		{
-			name:         "missing environment credential",
-			prompt:       basePrompt,
-			profile:      promptexec.ProfileInspection{ProfileID: "default-profile", APIKeyEnv: "PROMPT_API_KEY"},
-			lookupEnv:    func(string) (string, bool) { return "", false },
-			wantCategory: promptexec.MissingCredential,
-		},
 	}
 	for _, test := range tests {
 		t.Run(test.name, func(t *testing.T) {
 			executor := &inspectionExecutor{prompt: test.prompt, profiles: map[string]promptexec.ProfileInspection{"default-profile": test.profile}}
-			_, err := InspectPromptExecution(context.Background(), PromptInspectionRequest{Resolved: resolved, Executor: executor, LookupEnv: test.lookupEnv})
+			_, err := InspectPromptExecution(context.Background(), PromptInspectionRequest{Resolved: resolved, Executor: executor})
 			if err == nil || promptexec.CategoryOf(err) != test.wantCategory {
 				t.Fatalf("error/category = %v/%q, want %q", err, promptexec.CategoryOf(err), test.wantCategory)
 			}
@@ -265,13 +257,12 @@ func TestInspectComparisonExecutionStopsAtFirstProfileFailure(t *testing.T) {
 		prompt: validPromptInspection(resolved.Definition),
 		profiles: map[string]promptexec.ProfileInspection{
 			"weather-light": {ProfileID: "weather-light", BackendID: "local", ModelName: "light-model"},
-			"missing-key":   {ProfileID: "missing-key", APIKeyEnv: "PROMPT_API_KEY"},
+			"missing-key":   {ProfileID: "missing-key", CredentialRequired: true},
 			"weather-deep":  {ProfileID: "weather-deep", BackendID: "cloud", ModelName: "deep-model"},
 		},
 	}
 	result, err := InspectComparisonExecution(context.Background(), ComparisonInspectionRequest{
 		Resolved: resolved, ProfileIDs: []string{"weather-light", "missing-key", "weather-deep"}, Executor: executor,
-		LookupEnv: func(string) (string, bool) { return "", false },
 	})
 	if err == nil || promptexec.CategoryOf(err) != promptexec.MissingCredential {
 		t.Fatalf("error/category = %v/%q, want missing credential", err, promptexec.CategoryOf(err))
@@ -359,7 +350,7 @@ func validPromptInspection(definition report.Definition) promptexec.PromptInspec
 	return promptexec.PromptInspection{
 		PromptID: definition.PromptID, PromptVersion: definition.PromptVersion, PromptHash: "prompt-hash", DefaultProfileID: "default-profile",
 		Inputs: []promptexec.InputDefinition{{Name: "data_package", Required: true, ContentType: "application/yaml"}},
-		Output: promptexec.OutputContract{Format: "json", ValidationMode: "json_schema", SchemaPath: definition.GeneratedTextSchemaID + ".generated_text.schema.json"},
+		Output: promptexec.OutputContract{Format: "json", ValidationMode: "json_schema", SchemaPath: definition.GeneratedTextSchemaID + ".generated_text.schema.json", RepairAttempts: definition.GeneratedTextRepairAttempts},
 	}
 }
 

internal/app/prompt_profile_integration_test.go

@@ -2,8 +2,10 @@ package app_test
 
 import (
 	"context"
+	"fmt"
 	"os"
 	"path/filepath"
+	"strings"
 	"testing"
 	"time"
 
@@ -14,14 +16,12 @@ import (
 )
 
 func TestPromptInspectionResolvesEmbeddedAndOverriddenProfilesOffline(t *testing.T) {
-	lookupEnv := func(string) (string, bool) { return "test-key", true }
 	inspect := func(t *testing.T, adapter *promptkitadapter.Adapter, id report.ID, profile string, wantID string, wantBackend string, wantModel string) {
 		t.Helper()
 		result, err := app.InspectPromptExecution(context.Background(), app.PromptInspectionRequest{
 			Resolved:  resolvedPromptProfile(t, id),
 			Executor:  adapter,
 			Promptkit: config.PromptkitConfig{Profile: profile},
-			LookupEnv: lookupEnv,
 		})
 		if err != nil {
 			t.Fatalf("InspectPromptExecution() error = %v", err)
@@ -50,6 +50,55 @@ model: local-weather
 	inspect(t, override, report.Hourly, "", "weather-light", "openrouter", "local-weather")
 }
 
+func TestPromptInspectionAcceptsMaintainedEndpointOnlyProfile(t *testing.T) {
+	adapter, err := promptkitadapter.New(promptkitadapter.Config{ProfileFile: filepath.Join("..", "..", "examples", "weather-light-local-profile.yml")})
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	result, err := app.InspectPromptExecution(context.Background(), app.PromptInspectionRequest{
+		Resolved: resolvedPromptProfile(t, report.Hourly), Executor: adapter,
+	})
+	if err != nil {
+		t.Fatalf("InspectPromptExecution() error = %v", err)
+	}
+	if result.ProfileID != "weather-light" || result.BackendID != "" || result.ModelName != "weather-local" {
+		t.Fatalf("inspection = %#v", result)
+	}
+	if strings.Contains(fmt.Sprintf("%#v", result), "127.0.0.1") {
+		t.Fatalf("inspection leaks endpoint: %#v", result)
+	}
+}
+
+func TestPromptInspectionSupportsRakestrawhomeProfileOffline(t *testing.T) {
+	adapter, err := promptkitadapter.New(promptkitadapter.Config{})
+	if err != nil {
+		t.Fatalf("New() error = %v", err)
+	}
+	prompt, err := app.InspectPromptExecution(context.Background(), app.PromptInspectionRequest{
+		Resolved:  resolvedPromptProfile(t, report.Hourly),
+		Executor:  adapter,
+		Promptkit: config.PromptkitConfig{Profile: "rakestrawhome-gemma-4-31b"},
+	})
+	if err != nil {
+		t.Fatalf("InspectPromptExecution() error = %v", err)
+	}
+	if prompt.ProfileID != "rakestrawhome-gemma-4-31b" || prompt.BackendID != "rakestrawhome" || prompt.ModelName == "" {
+		t.Fatalf("prompt inspection = %#v", prompt)
+	}
+
+	comparison, err := app.InspectComparisonExecution(context.Background(), app.ComparisonInspectionRequest{
+		Resolved:   resolvedPromptProfile(t, report.Hourly),
+		ProfileIDs: []string{"rakestrawhome-gemma-4-31b", "weather-deep"},
+		Executor:   adapter,
+	})
+	if err != nil {
+		t.Fatalf("InspectComparisonExecution() error = %v", err)
+	}
+	if len(comparison.Profiles) != 2 || comparison.Profiles[0].ProfileID != "rakestrawhome-gemma-4-31b" || comparison.Profiles[0].BackendID != "rakestrawhome" || comparison.Profiles[0].ModelName == "" {
+		t.Fatalf("comparison inspection = %#v", comparison)
+	}
+}
+
 func resolvedPromptProfile(t *testing.T, id report.ID) report.Resolved {
 	t.Helper()
 	now := time.Date(2026, 5, 29, 12, 0, 0, 0, time.UTC)

internal/cli/comparison_test.go

@@ -355,7 +355,7 @@ func comparisonResult(outputDirectory string, results []app.ComparisonProfileRes
 	started := time.Date(2026, 5, 29, 13, 30, 0, 0, time.UTC)
 	result := &app.ComparisonResult{
 		ComparisonID: "comparison_run-123", ReportID: "daily", ReportName: "Daily Report",
-		PromptID: "weather.daily_generated_text", PromptVersion: "2.0.0", PromptHash: strings.Repeat("a", 64),
+		PromptID: "weather.daily_generated_text", PromptVersion: "2.1.0", PromptHash: strings.Repeat("a", 64),
 		StartedAt: started, FinishedAt: started.Add(time.Minute), Timezone: "America/Chicago",
 		ValidPeriod: timeutil.Period{Start: started, End: started.Add(24 * time.Hour)}, OutputDirectory: outputDirectory,
 		ManifestPath: filepath.Join(outputDirectory, "comparison.json"), DataPackagePath: filepath.Join(outputDirectory, "data-package.yml"),

internal/cli/generate_test.go

@@ -132,7 +132,7 @@ func actionConfigPath(t *testing.T) string {
 func generatedReportResult() *app.ReportResult {
 	generatedAt := time.Date(2026, 5, 29, 13, 30, 0, 0, time.UTC)
 	return &app.ReportResult{
-		ReportID: report.Daily, ReportName: "Daily Report", PromptID: "weather.daily_generated_text", PromptVersion: "2.0.0",
+		ReportID: report.Daily, ReportName: "Daily Report", PromptID: "weather.daily_generated_text", PromptVersion: "2.1.0",
 		RunID: "daily-20260529", GeneratedAt: generatedAt, Timezone: "America/Chicago",
 		ValidPeriod: timeutil.Period{Start: generatedAt, End: generatedAt.Add(24 * time.Hour)},
 		ProfileID:   "weather-light", BackendID: "local", ModelName: "weather-model", ValidationStatus: promptexec.ValidationPassed,

internal/cli/result.go

@@ -43,6 +43,7 @@ type generateSummary struct {
 	ModelName        string                       `json:"modelName,omitempty"`
 	SourceWarnings   []weatherdata.SourceWarning  `json:"sourceWarnings,omitempty"`
 	ValidationStatus string                       `json:"validationStatus,omitempty"`
+	RepairAttempts   *int                         `json:"repairAttempts,omitempty"`
 	Notification     *generateNotificationSummary `json:"notification,omitempty"`
 	Error            string                       `json:"error,omitempty"`
 }
@@ -106,6 +107,7 @@ type comparisonProfileSummary struct {
 	ModelName        string                `json:"modelName"`
 	Status           string                `json:"status"`
 	ValidationStatus string                `json:"validationStatus,omitempty"`
+	RepairAttempts   *int                  `json:"repairAttempts,omitempty"`
 	ReportPath       string                `json:"reportPath,omitempty"`
 	LLMDebugPath     string                `json:"llmDebugPath,omitempty"`
 	Error            *comparison.SafeError `json:"error,omitempty"`
@@ -129,6 +131,10 @@ func newGenerateSummary(result *app.ReportResult, err error) generateSummary {
 	summary.ProfileID, summary.BackendID, summary.ModelName = result.ProfileID, result.BackendID, result.ModelName
 	summary.SourceWarnings = append([]weatherdata.SourceWarning(nil), result.SourceWarnings...)
 	summary.ValidationStatus = string(result.ValidationStatus)
+	if result.RepairAttempts != nil {
+		value := *result.RepairAttempts
+		summary.RepairAttempts = &value
+	}
 	summary.OutputPath = result.OutputPath
 	summary.LLMDebugPath = result.LLMDebugPath
 	summary.Notification = newGenerateNotificationSummary(result.Notification)
@@ -162,6 +168,11 @@ func newGenerateNotificationSummary(result *app.NotificationResult) *generateNot
 	return summary
 }
 
+func repairAttemptsCopy(value int) *int {
+	copy := value
+	return &copy
+}
+
 func newBatchSummary(result *app.BatchResult, err error) batchSummary {
 	summary := batchSummary{Command: commandRun}
 	if result == nil {
@@ -203,6 +214,9 @@ func newComparisonSummary(result *app.ComparisonResult, err error) comparisonSum
 			Status: profile.Status, ValidationStatus: string(profile.ValidationStatus), ReportPath: profile.ReportPath,
 			LLMDebugPath: profile.LLMDebugPath, Error: profile.Error,
 		})
+		if profile.RepairAttempts != nil {
+			summary.Results[len(summary.Results)-1].RepairAttempts = repairAttemptsCopy(*profile.RepairAttempts)
+		}
 	}
 	summary.Status = comparisonSummaryStatus(result, err)
 	if err != nil {

internal/cli/result_test.go

@@ -19,7 +19,7 @@ import (
 
 func TestGenerateSummaryUsesActiveResultFields(t *testing.T) {
 	generatedAt := time.Date(2026, 5, 29, 13, 30, 0, 0, time.UTC)
-	summary := newGenerateSummary(&app.ReportResult{ReportID: report.Daily, ReportName: "Daily Report", PromptID: "weather.daily_generated_text", PromptVersion: "2.0.0", RunID: "run-123", GeneratedAt: generatedAt, Timezone: "America/Chicago", ValidPeriod: timeutil.Period{Start: generatedAt, End: generatedAt.Add(24 * time.Hour)}, ProfileID: "weather-balanced", BackendID: "openrouter", ModelName: "model", SourceWarnings: []weatherdata.SourceWarning{{Source: "alerts", Message: "source unavailable"}}, ValidationStatus: promptexec.ValidationPassed, OutputPath: "/reports/daily.md"}, nil)
+	summary := newGenerateSummary(&app.ReportResult{ReportID: report.Daily, ReportName: "Daily Report", PromptID: "weather.daily_generated_text", PromptVersion: "2.1.0", RunID: "run-123", GeneratedAt: generatedAt, Timezone: "America/Chicago", ValidPeriod: timeutil.Period{Start: generatedAt, End: generatedAt.Add(24 * time.Hour)}, ProfileID: "weather-balanced", BackendID: "openrouter", ModelName: "model", SourceWarnings: []weatherdata.SourceWarning{{Source: "alerts", Message: "source unavailable"}}, ValidationStatus: promptexec.ValidationPassed, OutputPath: "/reports/daily.md"}, nil)
 	if summary.OutputPath == "" || summary.ProfileID == "" || summary.ValidationStatus != string(promptexec.ValidationPassed) || len(summary.SourceWarnings) != 1 {
 		t.Fatalf("summary = %#v", summary)
 	}
@@ -39,7 +39,7 @@ func TestComparisonSummaryUsesLockedOrderAndSafeFields(t *testing.T) {
 	profileFailure := comparison.NewSafeError("generation", "execute prompt failed")
 	result := &app.ComparisonResult{
 		ComparisonID: "comparison_run-123", ReportID: report.Daily, ReportName: "Daily Report",
-		PromptID: "weather.daily_generated_text", PromptVersion: "2.0.0", PromptHash: strings.Repeat("a", 64),
+		PromptID: "weather.daily_generated_text", PromptVersion: "2.1.0", PromptHash: strings.Repeat("a", 64),
 		StartedAt: started, FinishedAt: started.Add(time.Minute), Timezone: "America/Chicago",
 		ValidPeriod: timeutil.Period{Start: started, End: started.Add(24 * time.Hour)}, OutputDirectory: "/reports/comparison-daily",
 		ManifestPath: "/reports/comparison-daily/comparison.json", DataPackagePath: "/reports/comparison-daily/data-package.yml",
@@ -113,7 +113,7 @@ func TestComparisonSummaryClassifiesCompleteAndAllFailedResults(t *testing.T) {
 	started := time.Date(2026, 5, 29, 13, 30, 0, 0, time.UTC)
 	failure := comparison.NewSafeError("generation", "execute prompt failed")
 	complete := &app.ComparisonResult{
-		ComparisonID: "comparison_run-123", ReportID: report.Daily, PromptID: "weather.daily_generated_text", PromptVersion: "2.0.0", PromptHash: strings.Repeat("a", 64),
+		ComparisonID: "comparison_run-123", ReportID: report.Daily, PromptID: "weather.daily_generated_text", PromptVersion: "2.1.0", PromptHash: strings.Repeat("a", 64),
 		StartedAt: started, FinishedAt: started, Timezone: "America/Chicago", ValidPeriod: timeutil.Period{Start: started, End: started.Add(time.Hour)},
 		OutputDirectory: "/reports/comparison-daily", ManifestPath: "/reports/comparison-daily/comparison.json", DataPackagePath: "/reports/comparison-daily/data-package.yml",
 		Total: 2, Succeeded: 2,

internal/comparison/comparison.go

@@ -15,7 +15,7 @@ import (
 
 const (
 	// SchemaVersion identifies the supported comparison manifest schema.
-	SchemaVersion = "weatherreporter.comparison.v1"
+	SchemaVersion = "weatherreporter.comparison.v2"
 
 	// ManifestFilename is the canonical name of a comparison manifest.
 	ManifestFilename = "comparison.json"
@@ -71,6 +71,7 @@ type Result struct {
 	ModelName        string     `json:"modelName"`
 	Status           string     `json:"status"`
 	ValidationStatus string     `json:"validationStatus,omitempty"`
+	RepairAttempts   *int       `json:"repairAttempts,omitempty"`
 	ReportPath       string     `json:"reportPath,omitempty"`
 	Error            *SafeError `json:"error,omitempty"`
 }
@@ -276,11 +277,14 @@ func (manifest Manifest) Validate() error {
 		if strings.TrimSpace(result.ModelName) == "" {
 			return fmt.Errorf("result %d has a blank model name", result.Position)
 		}
+		if result.RepairAttempts != nil && *result.RepairAttempts < 0 {
+			return fmt.Errorf("result %d has negative repair attempts", result.Position)
+		}
 
 		switch result.Status {
 		case StatusSucceeded:
 			succeeded++
-			if result.ValidationStatus != "passed" {
+			if result.ValidationStatus != "passed" || result.RepairAttempts == nil {
 				return fmt.Errorf("successful result %d did not pass validation", result.Position)
 			}
 			expectedPath, err := ReportFilename(result.Position, manifest.Total, result.ProfileID)
@@ -302,6 +306,9 @@ func (manifest Manifest) Validate() error {
 			if err := result.Error.validate(); err != nil {
 				return fmt.Errorf("failed result %d: %w", result.Position, err)
 			}
+			if result.ValidationStatus != "" && result.RepairAttempts == nil {
+				return fmt.Errorf("failed result %d has validation without repair provenance", result.Position)
+			}
 		default:
 			return fmt.Errorf("result %d has unsupported status %q", result.Position, result.Status)
 		}
@@ -324,7 +331,7 @@ func (safeError SafeError) validate() error {
 }
 
 func isValidationStatus(status string) bool {
-	return status == "" || status == "failed" || status == "skipped"
+	return status == "" || status == "passed" || status == "failed" || status == "skipped"
 }
 
 func isSHA256(value string) bool {

internal/comparison/comparison_test.go

@@ -118,7 +118,7 @@ func TestManifestEncodingAndRoundTrip(t *testing.T) {
 		t.Fatalf("EncodeManifest() error = %v", err)
 	}
 	want := "{\n" +
-		"  \"schemaVersion\": \"weatherreporter.comparison.v1\",\n" +
+		"  \"schemaVersion\": \"weatherreporter.comparison.v2\",\n" +
 		"  \"comparisonId\": \"comparison_daily-2026-08-24\",\n" +
 		"  \"startedAt\": \"2026-08-24T12:00:00Z\",\n" +
 		"  \"finishedAt\": \"2026-08-24T12:01:00Z\",\n" +
@@ -146,6 +146,7 @@ func TestManifestEncodingAndRoundTrip(t *testing.T) {
 		"      \"modelName\": \"gpt-5-mini\",\n" +
 		"      \"status\": \"succeeded\",\n" +
 		"      \"validationStatus\": \"passed\",\n" +
+		"      \"repairAttempts\": 0,\n" +
 		"      \"reportPath\": \"01-weather-light.md\"\n" +
 		"    },\n" +
 		"    {\n" +
@@ -201,6 +202,9 @@ func TestManifestValidateRejectsInvariants(t *testing.T) {
 		{name: "successful result without passed validation", mutate: func(manifest *Manifest) { manifest.Results[0].ValidationStatus = "failed" }},
 		{name: "failed result with report", mutate: func(manifest *Manifest) { manifest.Results[1].ReportPath = "02-weather-deep.md" }},
 		{name: "failed result without error", mutate: func(manifest *Manifest) { manifest.Results[1].Error = nil }},
+		{name: "failed passed result without repair provenance", mutate: func(manifest *Manifest) {
+			manifest.Results[1].ValidationStatus = "passed"
+		}},
 		{name: "traversal report path", mutate: func(manifest *Manifest) { manifest.Results[0].ReportPath = "../report.md" }},
 		{name: "duplicate report path", mutate: func(manifest *Manifest) {
 			manifest.Results[1] = Result{Position: 2, ProfileID: "weather-deep", ModelName: "gpt-5", Status: StatusSucceeded, ValidationStatus: "passed", ReportPath: manifest.Results[0].ReportPath}
@@ -222,6 +226,17 @@ func TestManifestValidateRejectsInvariants(t *testing.T) {
 	}
 }
 
+func TestManifestValidateAcceptsPostValidationFailure(t *testing.T) {
+	t.Parallel()
+
+	manifest := validManifest()
+	manifest.Results[1].ValidationStatus = "passed"
+	manifest.Results[1].RepairAttempts = intPtr(0)
+	if err := manifest.Validate(); err != nil {
+		t.Fatalf("Manifest.Validate() rejected post-validation failure: %v", err)
+	}
+}
+
 func TestLogicalBundleValidate(t *testing.T) {
 	t.Parallel()
 
@@ -309,6 +324,7 @@ func validManifest() Manifest {
 				ModelName:        "gpt-5-mini",
 				Status:           StatusSucceeded,
 				ValidationStatus: "passed",
+				RepairAttempts:   intPtr(0),
 				ReportPath:       "01-weather-light.md",
 			},
 			{
@@ -321,3 +337,8 @@ func validManifest() Manifest {
 		},
 	}
 }
+
+func intPtr(value int) *int {
+	copy := value
+	return &copy
+}

internal/comparison/publish.go

@@ -388,6 +388,7 @@ var resultFields = map[string]jsonValueValidator{
 	"modelName":        nil,
 	"status":           nil,
 	"validationStatus": nil,
+	"repairAttempts":   validateRepairAttemptsJSON,
 	"reportPath":       nil,
 	"error":            validateSafeErrorJSON,
 }
@@ -430,7 +431,7 @@ func validateResultsJSON(decoder *json.Decoder) error {
 		return fmt.Errorf("results must be an array")
 	}
 	for decoder.More() {
-		if err := validateJSONObject(decoder, resultFields); err != nil {
+		if err := validateOrderedJSONObject(decoder, resultFields, []string{"position", "profileId", "backendId", "modelName", "status", "validationStatus", "repairAttempts", "reportPath", "error"}); err != nil {
 			return err
 		}
 	}
@@ -444,6 +445,65 @@ func validateResultsJSON(decoder *json.Decoder) error {
 	return nil
 }
 
+func validateOrderedJSONObject(decoder *json.Decoder, fields map[string]jsonValueValidator, order []string) error {
+	token, err := decoder.Token()
+	if err != nil {
+		return err
+	}
+	if delimiter, ok := token.(json.Delim); !ok || delimiter != '{' {
+		return fmt.Errorf("manifest value must be an object")
+	}
+	seen := make(map[string]struct{}, len(fields))
+	last := -1
+	for decoder.More() {
+		token, err := decoder.Token()
+		if err != nil {
+			return err
+		}
+		name, ok := token.(string)
+		if !ok {
+			return fmt.Errorf("manifest field name is invalid")
+		}
+		validator, known := fields[name]
+		if !known {
+			return fmt.Errorf("manifest field %q is not canonical", name)
+		}
+		if _, duplicate := seen[name]; duplicate {
+			return fmt.Errorf("manifest field %q is duplicated", name)
+		}
+		position := -1
+		for index, candidate := range order {
+			if candidate == name {
+				position = index
+				break
+			}
+		}
+		if position <= last {
+			return fmt.Errorf("manifest field %q is out of canonical order", name)
+		}
+		seen[name] = struct{}{}
+		last = position
+		if validator == nil {
+			var value json.RawMessage
+			if err := decoder.Decode(&value); err != nil {
+				return err
+			}
+			continue
+		}
+		if err := validator(decoder); err != nil {
+			return err
+		}
+	}
+	token, err = decoder.Token()
+	if err != nil {
+		return err
+	}
+	if delimiter, ok := token.(json.Delim); !ok || delimiter != '}' {
+		return fmt.Errorf("manifest object has an invalid terminator")
+	}
+	return nil
+}
+
 func validateSafeErrorJSON(decoder *json.Decoder) error {
 	token, err := decoder.Token()
 	if err != nil {
@@ -458,6 +518,24 @@ func validateSafeErrorJSON(decoder *json.Decoder) error {
 	return validateJSONObjectBody(decoder, safeErrorFields)
 }
 
+func validateRepairAttemptsJSON(decoder *json.Decoder) error {
+	var raw json.RawMessage
+	if err := decoder.Decode(&raw); err != nil {
+		return err
+	}
+	if bytes.Equal(bytes.TrimSpace(raw), []byte("null")) {
+		return fmt.Errorf("repairAttempts must be an integer")
+	}
+	var value int
+	if err := json.Unmarshal(raw, &value); err != nil {
+		return err
+	}
+	if value < 0 {
+		return fmt.Errorf("repairAttempts must not be negative")
+	}
+	return nil
+}
+
 func validateJSONObject(decoder *json.Decoder, fields map[string]jsonValueValidator) error {
 	token, err := decoder.Token()
 	if err != nil {

internal/comparison/publish_test.go

@@ -1,6 +1,7 @@
 package comparison
 
 import (
+	"bytes"
 	"context"
 	"encoding/json"
 	"errors"
@@ -146,6 +147,23 @@ func TestRecognizeBundleRejectsUnsafeAndMalformedContents(t *testing.T) {
 			t.Helper()
 			appendManifestField(t, directory, `"SchemaVersion": "weatherreporter.comparison.v1"`)
 		}},
+		{name: "null repair attempts", mutate: func(t *testing.T, directory string) {
+			t.Helper()
+			path := filepath.Join(directory, ManifestFilename)
+			data, err := os.ReadFile(path)
+			if err != nil {
+				t.Fatal(err)
+			}
+			before := []byte("      \"status\": \"failed\",\n      \"error\": {")
+			after := []byte("      \"status\": \"failed\",\n      \"repairAttempts\": null,\n      \"error\": {")
+			data = bytes.Replace(data, before, after, 1)
+			if bytes.Equal(data, readFile(t, path)) {
+				t.Fatal("failed to add null repairAttempts")
+			}
+			if err := os.WriteFile(path, data, 0o600); err != nil {
+				t.Fatal(err)
+			}
+		}},
 		{name: "traversal report path", mutate: func(t *testing.T, directory string) {
 			t.Helper()
 			path := filepath.Join(directory, ManifestFilename)
@@ -202,6 +220,30 @@ func TestRecognizeBundleRejectsUnsafeAndMalformedContents(t *testing.T) {
 	}
 }
 
+func TestValidateRepairAttemptsJSONRejectsNonIntegerValues(t *testing.T) {
+	t.Parallel()
+
+	for _, value := range []string{"null", "-1", "1.5", `"1"`, "9223372036854775808", "{"} {
+		value := value
+		t.Run(value, func(t *testing.T) {
+			t.Parallel()
+			if err := validateRepairAttemptsJSON(json.NewDecoder(strings.NewReader(value))); err == nil {
+				t.Fatalf("validateRepairAttemptsJSON(%q) error = nil", value)
+			}
+		})
+	}
+}
+
+func TestValidateRepairAttemptsJSONAcceptsNonnegativeIntegers(t *testing.T) {
+	t.Parallel()
+
+	for _, value := range []string{"0", "1"} {
+		if err := validateRepairAttemptsJSON(json.NewDecoder(strings.NewReader(value))); err != nil {
+			t.Fatalf("validateRepairAttemptsJSON(%q) error = %v", value, err)
+		}
+	}
+}
+
 func appendManifestField(t *testing.T, directory, field string) {
 	t.Helper()
 	path := filepath.Join(directory, ManifestFilename)
@@ -1077,7 +1119,7 @@ func testBundleWithReports(t *testing.T, count int) LogicalBundle {
 		}
 		manifest.Results[i] = Result{
 			Position: position, ProfileID: profileID, ModelName: "gpt-5-mini",
-			Status: StatusSucceeded, ValidationStatus: "passed", ReportPath: path,
+			Status: StatusSucceeded, ValidationStatus: "passed", RepairAttempts: intPtr(0), ReportPath: path,
 		}
 		reports[i] = BundleReport{Position: position, Path: path, Markdown: []byte("# Daily\n")}
 	}

internal/promptassets/assets/profiles/weather-balanced.yml

@@ -1,6 +1,2 @@
 id: weather-balanced
-backend: openrouter
-model: "~google/gemini-flash-latest"
-reasoning_effort: high
-timeout_seconds: 240
-service_tier: flex
+base_profile: gemini-flash-latest

internal/promptassets/assets/profiles/weather-deep.yml

@@ -1,6 +1,2 @@
 id: weather-deep
-backend: openrouter
-model: "~anthropic/claude-sonnet-latest"
-reasoning_effort: high
-timeout_seconds: 240
-service_tier: flex
+base_profile: claude-sonnet-latest

internal/promptassets/assets/profiles/weather-light.yml

@@ -1,5 +1,2 @@
 id: weather-light
-backend: openrouter
-model: deepseek/deepseek-v4-flash
-timeout_seconds: 180
-service_tier: flex
+base_profile: deepseek-4-flash

internal/promptassets/assets/prompts/daily/daily_generated_text.yml

@@ -1,5 +1,5 @@
 id: weather.daily_generated_text
-version: "2.0.0"
+version: "2.1.0"
 default_profile: weather-balanced
 description: Daily weather report analysis prompt.
 inputs:
@@ -21,3 +21,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: daily.generated_text.schema.json
+  repair_attempts: 1

internal/promptassets/assets/prompts/hourly/hourly_generated_text.yml

@@ -1,5 +1,5 @@
 id: weather.hourly_generated_text
-version: "2.0.0"
+version: "2.1.0"
 default_profile: weather-light
 description: Hourly weather report analysis prompt.
 inputs:
@@ -21,3 +21,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: hourly.generated_text.schema.json
+  repair_attempts: 1

internal/promptassets/assets/prompts/today/today_generated_text.yml

@@ -1,5 +1,5 @@
 id: weather.today_generated_text
-version: "2.0.0"
+version: "2.1.0"
 default_profile: weather-balanced
 description: Today's weather report analysis prompt.
 inputs:
@@ -21,3 +21,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: today.generated_text.schema.json
+  repair_attempts: 1

internal/promptassets/assets/prompts/tomorrow/tomorrow_generated_text.yml

@@ -1,5 +1,5 @@
 id: weather.tomorrow_generated_text
-version: "2.0.0"
+version: "2.1.0"
 default_profile: weather-balanced
 description: Tomorrow's weather report analysis prompt.
 inputs:
@@ -21,3 +21,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: tomorrow.generated_text.schema.json
+  repair_attempts: 1

internal/promptassets/promptassets_test.go

@@ -77,8 +77,8 @@ func TestPromptAssetsDeclareTheFourGeneratedTextPrompts(t *testing.T) {
 			if err := yaml.Unmarshal(data, &definition); err != nil {
 				t.Fatalf("decode prompt definition: %v", err)
 			}
-			if definition.ID != tc.id || definition.Version != "2.0.0" || definition.DefaultProfile != tc.profile {
-				t.Fatalf("definition = %#v, want %s version 2.0.0 and profile %s", definition, tc.id, tc.profile)
+			if definition.ID != tc.id || definition.Version != "2.1.0" || definition.DefaultProfile != tc.profile {
+				t.Fatalf("definition = %#v, want %s version 2.1.0 and profile %s", definition, tc.id, tc.profile)
 			}
 			sharedInstruction := false
 			for _, message := range definition.Messages {
@@ -92,8 +92,8 @@ func TestPromptAssetsDeclareTheFourGeneratedTextPrompts(t *testing.T) {
 			if len(definition.Inputs) != 1 || definition.Inputs[0].Name != "data_package" || !definition.Inputs[0].Required || definition.Inputs[0].ContentType != "application/yaml" {
 				t.Fatalf("inputs = %#v, want one required YAML data_package", definition.Inputs)
 			}
-			if definition.Output.Format != "json" || definition.Output.ValidationMode != "json_schema" || definition.Output.SchemaPath != tc.schemaID+".generated_text.schema.json" || definition.Output.RepairAttempts != nil {
-				t.Fatalf("output = %#v, want JSON schema output without repair attempts", definition.Output)
+			if definition.Output.Format != "json" || definition.Output.ValidationMode != "json_schema" || definition.Output.SchemaPath != tc.schemaID+".generated_text.schema.json" || definition.Output.RepairAttempts == nil || *definition.Output.RepairAttempts != 1 {
+				t.Fatalf("output = %#v, want JSON schema output with one repair attempt", definition.Output)
 			}
 			if _, err := promptassets.Schema(tc.schemaID); err != nil {
 				t.Fatalf("Schema(%q) error = %v", tc.schemaID, err)
@@ -286,11 +286,11 @@ func TestPromptkitInspectsEmbeddedPromptsOffline(t *testing.T) {
 		{"weather.hourly_generated_text", "weather-light", "deepseek/deepseek-v4-flash"},
 	} {
 		t.Run(want.id, func(t *testing.T) {
-			inspection, err := engine.InspectPrompt(context.Background(), want.id, "2.0.0")
+			inspection, err := engine.InspectPrompt(context.Background(), want.id, "2.1.0")
 			if err != nil {
 				t.Fatalf("InspectPrompt() error = %v", err)
 			}
-			if inspection.PromptID != want.id || inspection.PromptVersion != "2.0.0" || inspection.DefaultProfileID != want.profile {
+			if inspection.PromptID != want.id || inspection.PromptVersion != "2.1.0" || inspection.DefaultProfileID != want.profile || inspection.OutputContract.RepairAttempts != 1 {
 				t.Fatalf("inspection = %#v", inspection)
 			}
 			profile, err := engine.InspectProfile(context.Background(), inspection.DefaultProfileID)
@@ -361,6 +361,29 @@ func TestEmbeddedProfilesAreCompleteAndInspectable(t *testing.T) {
 	}
 }
 
+func TestEmbeddedProfilesAreMinimalBaseAliases(t *testing.T) {
+	wantBases := map[string]string{
+		"weather-balanced.yml": "gemini-flash-latest",
+		"weather-deep.yml":     "claude-sonnet-latest",
+		"weather-light.yml":    "deepseek-4-flash",
+	}
+	for path, wantBase := range wantBases {
+		t.Run(path, func(t *testing.T) {
+			data, err := fs.ReadFile(promptassets.ProfileFS(), path)
+			if err != nil {
+				t.Fatalf("read profile: %v", err)
+			}
+			var definition map[string]string
+			if err := yaml.Unmarshal(data, &definition); err != nil {
+				t.Fatalf("unmarshal profile: %v", err)
+			}
+			if definition["id"] != strings.TrimSuffix(path, ".yml") || definition["base_profile"] != wantBase || len(definition) != 2 {
+				t.Fatalf("profile definition = %#v, want only its id and base profile %q", definition, wantBase)
+			}
+		})
+	}
+}
+
 func TestEmbeddedProfilesExcludeUnsafeOrIncidentalSettings(t *testing.T) {
 	forbidden := []string{"endpoint:", "api_key", "credential", "temperature:", "top_p:", "max_tokens:"}
 	if err := fs.WalkDir(promptassets.ProfileFS(), ".", func(path string, entry fs.DirEntry, err error) error {
@@ -391,7 +414,7 @@ func TestPromptAssetsExcludeRetiredRuntimeSettings(t *testing.T) {
 		if err != nil {
 			return err
 		}
-		for _, unwanted := range []string{"local-heavy", "pipeline-weather/", "application/json", "repair_attempts:", "weather.daily_report"} {
+		for _, unwanted := range []string{"local-heavy", "pipeline-weather/", "application/json", "weather.daily_report"} {
 			if strings.Contains(string(data), unwanted) {
 				t.Fatalf("%s contains retired runtime setting %q", path, unwanted)
 			}

internal/promptdebug/debug_writer.go

@@ -21,8 +21,9 @@ import (
 var ErrSecureCaptureUnsupported = errors.New("secure prompt debug capture is unavailable on this platform")
 
 const (
-	promptPreparationDebugSchemaVersion = "weatherreporter.prompt_preparation_debug.v2"
-	promptExecutionDebugSchemaVersion   = "weatherreporter.prompt_execution_debug.v2"
+	promptPreparationDebugSchemaVersion = "weatherreporter.prompt_preparation_debug.v3"
+	promptExecutionDebugSchemaVersion   = "weatherreporter.prompt_execution_debug.v3"
+	promptFailureDebugSchemaVersion     = "weatherreporter.prompt_failure_debug.v1"
 	debugDirectoryMode                  = 0o700
 	debugFileMode                       = 0o600
 )
@@ -53,6 +54,7 @@ type PromptDebugOutput struct {
 	Format         string `json:"format"`
 	ValidationMode string `json:"validationMode"`
 	SchemaPath     string `json:"schemaPath"`
+	RepairAttempts int    `json:"repairAttempts"`
 }
 
 // PromptDebugPreparation is the explicit, content-safe mapping of preparation
@@ -100,9 +102,27 @@ type PromptDebugValidation struct {
 	Status         string   `json:"status"`
 	Mode           string   `json:"mode"`
 	SchemaPath     string   `json:"schemaPath"`
+	RepairAttempts int      `json:"repairAttempts"`
 	Diagnostics    []string `json:"diagnostics,omitempty"`
 }
 
+// PromptFailureDebugArtifact records provider detail only in explicit debug storage.
+type PromptFailureDebugArtifact struct {
+	SchemaVersion string        `json:"schemaVersion"`
+	ReportID      report.ID     `json:"reportId"`
+	ValidDate     string        `json:"validDate"`
+	RunID         string        `json:"runId"`
+	Failure       PromptFailure `json:"failure"`
+}
+
+type PromptFailure struct {
+	Category        string `json:"category"`
+	StatusCode      int    `json:"statusCode,omitempty"`
+	ProviderCode    string `json:"providerCode,omitempty"`
+	ProviderType    string `json:"providerType,omitempty"`
+	ProviderMessage string `json:"providerMessage,omitempty"`
+}
+
 // PromptDebugExecution is the explicit mapping of execution provenance.
 type PromptDebugExecution struct {
 	RunID              string            `json:"runId"`
@@ -235,6 +255,24 @@ func (w *PromptDebugWriter) WriteExecution(ref PromptDebugRef, execution prompte
 	return directory, nil
 }
 
+// WriteFailure stores the project-owned structured provider failure.
+func (w *PromptDebugWriter) WriteFailure(ref PromptDebugRef, failure *promptexec.GenerationError) (string, error) {
+	if !w.Enabled() || failure == nil {
+		return "", nil
+	}
+	directory, secureDirectory, err := w.runDirectory(ref)
+	if err != nil {
+		return "", err
+	}
+	defer secureDirectory.Close()
+	artifact := PromptFailureDebugArtifact{SchemaVersion: promptFailureDebugSchemaVersion, ReportID: ref.ReportID, ValidDate: ref.ValidDate, RunID: ref.RunID,
+		Failure: PromptFailure{Category: string(failure.Category()), StatusCode: failure.StatusCode(), ProviderCode: failure.ProviderCode(), ProviderType: failure.ProviderType(), ProviderMessage: failure.ProviderMessage()}}
+	if err := secureDirectory.writeJSON("failure.json", artifact); err != nil {
+		return "", err
+	}
+	return directory, nil
+}
+
 func (w *PromptDebugWriter) runDirectory(ref PromptDebugRef) (string, *secureDirectory, error) {
 	if err := validatePromptDebugRef(ref); err != nil {
 		return "", nil, err
@@ -283,7 +321,7 @@ func promptDebugPreparation(value promptexec.Preparation) PromptDebugPreparation
 		PromptID: value.PromptID, PromptVersion: value.PromptVersion, PromptHash: value.PromptHash,
 		RenderedPromptHash: value.RenderedPromptHash, InputHashes: copyPromptDebugMap(value.InputHashes),
 		ProfileID: value.ProfileID, BackendID: value.BackendID, ModelName: value.ModelName,
-		Output:    PromptDebugOutput{Format: value.Output.Format, ValidationMode: value.Output.ValidationMode, SchemaPath: value.Output.SchemaPath},
+		Output:    PromptDebugOutput{Format: value.Output.Format, ValidationMode: value.Output.ValidationMode, SchemaPath: value.Output.SchemaPath, RepairAttempts: value.Output.RepairAttempts},
 		StartedAt: value.StartedAt, EndedAt: value.EndedAt, Duration: value.Duration,
 	}
 }
@@ -300,7 +338,7 @@ func promptDebugExecution(value promptexec.Execution) PromptDebugExecution {
 }
 
 func promptDebugValidation(value promptexec.Validation) PromptDebugValidation {
-	return PromptDebugValidation{Status: string(value.Status), Mode: value.Mode, SchemaPath: value.SchemaPath, Diagnostics: append([]string(nil), value.Diagnostics...)}
+	return PromptDebugValidation{Status: string(value.Status), Mode: value.Mode, SchemaPath: value.SchemaPath, RepairAttempts: value.RepairAttempts, Diagnostics: append([]string(nil), value.Diagnostics...)}
 }
 
 func promptDebugMessages(values []promptexec.RenderedMessage) []PromptDebugMessage {

internal/promptdebug/debug_writer_test.go

@@ -44,13 +44,13 @@ func TestPromptDebugWriterWritesIsolatedArtifacts(t *testing.T) {
 		t.Fatalf("WriteExecution() directory = %q, want %q", executionDir, preparationDir)
 	}
 	preparationData := readPromptDebugFile(t, filepath.Join(preparationDir, "preparation.json"))
-	for _, want := range []string{"weatherreporter.prompt_preparation_debug.v2", "Use the supplied weather facts.", `"type": "object"`, "https://llm.example.test", `"temperature": 0.2`} {
+	for _, want := range []string{"weatherreporter.prompt_preparation_debug.v3", "Use the supplied weather facts.", `"type": "object"`, "https://llm.example.test", `"temperature": 0.2`, `"repairAttempts": 0`} {
 		if !strings.Contains(string(preparationData), want) {
 			t.Fatalf("preparation debug artifact missing %q:\n%s", want, preparationData)
 		}
 	}
 	executionData := readPromptDebugFile(t, filepath.Join(executionDir, "execution.json"))
-	for _, want := range []string{"weatherreporter.prompt_execution_debug.v2", "Generated forecast prose.", "validation details", `"status": "passed"`} {
+	for _, want := range []string{"weatherreporter.prompt_execution_debug.v3", "Generated forecast prose.", "validation details", `"status": "passed"`, `"repairAttempts": 0`} {
 		if !strings.Contains(string(executionData), want) {
 			t.Fatalf("execution debug artifact missing %q:\n%s", want, executionData)
 		}
@@ -66,6 +66,25 @@ func TestPromptDebugWriterWritesIsolatedArtifacts(t *testing.T) {
 	assertPromptDebugMode(t, filepath.Join(executionDir, "execution.json"), debugFileMode)
 }
 
+func TestPromptDebugWriterWritesProviderFailureOnlyToDebugStore(t *testing.T) {
+	writer, err := NewPromptDebugWriter(filepath.Join(t.TempDir(), "operator-debug"))
+	if err != nil {
+		t.Fatalf("NewPromptDebugWriter() error = %v", err)
+	}
+	marker := "provider-private-marker"
+	directory, err := writer.WriteFailure(promptDebugRef(), promptexec.NewGenerationError(429, "rate_limit", "provider_error", marker, nil))
+	if err != nil {
+		t.Fatalf("WriteFailure() error = %v", err)
+	}
+	data := readPromptDebugFile(t, filepath.Join(directory, "failure.json"))
+	for _, want := range []string{"weatherreporter.prompt_failure_debug.v1", `"category": "generation"`, `"statusCode": 429`, marker} {
+		if !strings.Contains(string(data), want) {
+			t.Fatalf("failure artifact missing %q: %s", want, data)
+		}
+	}
+	assertPromptDebugMode(t, filepath.Join(directory, "failure.json"), debugFileMode)
+}
+
 func TestPromptDebugWriterProjectsProviderConfigurationSafely(t *testing.T) {
 	const marker = "private-debug-marker"
 	tests := []struct {
@@ -349,7 +368,7 @@ func promptDebugExecutionFixture() promptexec.Execution {
 		InputHashes: map[string]string{"data_package": "input-hash"}, ProfileID: "local", BackendID: "local", ModelName: "weather-model",
 		GeneratedHash: "generated-hash", Usage: promptexec.TokenUsage{PromptTokens: 10, CompletionTokens: 5, TotalTokens: 15},
 		StartedAt: startedAt, EndedAt: startedAt.Add(time.Second), Duration: time.Second,
-		Validation: promptexec.NewValidation(promptexec.ValidationPassed, "strict", "schemas/daily.json", []string{"validation details"}),
+		Validation: promptexec.NewValidation(promptexec.ValidationPassed, "strict", "schemas/daily.json", 0, []string{"validation details"}),
 		RawOutput:  []byte("Generated forecast prose."),
 		Debug:      &promptexec.ExecutionDebug{ValidationDiagnostics: []string{"validation details"}},
 	}

internal/promptexec/copy.go

@@ -78,3 +78,15 @@ func boundText(value string, limit int) string {
 	}
 	return value
 }
+
+func boundCodePoints(value string, limit int) string {
+	if limit <= 0 || value == "" {
+		return ""
+	}
+	value = strings.ToValidUTF8(value, "<22>")
+	runes := []rune(value)
+	if len(runes) <= limit {
+		return value
+	}
+	return string(runes[:limit])
+}

internal/promptexec/promptexec.go

@@ -51,6 +51,7 @@ type OutputContract struct {
 	Format         string
 	ValidationMode string
 	SchemaPath     string
+	RepairAttempts int
 }
 
 // ProfileInspection describes the safe, selected execution identity for one profile.
@@ -144,15 +145,17 @@ type Validation struct {
 	Status         ValidationStatus
 	Mode           string
 	SchemaPath     string
+	RepairAttempts int
 	Diagnostics    []string
 }
 
 // NewValidation returns a completed validation value with bounded diagnostics.
-func NewValidation(status ValidationStatus, mode string, schemaPath string, diagnostics []string) Validation {
+func NewValidation(status ValidationStatus, mode string, schemaPath string, repairAttempts int, diagnostics []string) Validation {
 	return Validation{
 		Status:         status,
 		Mode:           mode,
 		SchemaPath:     schemaPath,
+		RepairAttempts: repairAttempts,
 		Diagnostics:    boundDiagnostics(diagnostics),
 	}
 }
@@ -235,6 +238,86 @@ func (e *Error) Category() ErrorCategory {
 	return e.category
 }
 
+// GenerationError retains provider failure details for programmatic handling
+// without exposing them through routine formatting or serialization.
+type GenerationError struct {
+	statusCode      int
+	providerCode    string
+	providerType    string
+	providerMessage string
+	err             *Error
+}
+
+// NewGenerationError returns a classified generation failure with bounded
+// provider details. The dependency cause remains reachable through err only.
+func NewGenerationError(statusCode int, providerCode string, providerType string, providerMessage string, cause error) *GenerationError {
+	return &GenerationError{
+		statusCode:      statusCode,
+		providerCode:    boundCodePoints(providerCode, 256),
+		providerType:    boundCodePoints(providerType, 256),
+		providerMessage: boundCodePoints(providerMessage, 4096),
+		err:             NewError(Generation, "provider generation failed", cause),
+	}
+}
+
+// StatusCode returns the provider HTTP status when one was available.
+func (e *GenerationError) StatusCode() int {
+	if e == nil {
+		return 0
+	}
+	return e.statusCode
+}
+
+// ProviderCode returns the bounded provider error code.
+func (e *GenerationError) ProviderCode() string {
+	if e == nil {
+		return ""
+	}
+	return e.providerCode
+}
+
+// ProviderType returns the bounded provider error type.
+func (e *GenerationError) ProviderType() string {
+	if e == nil {
+		return ""
+	}
+	return e.providerType
+}
+
+// ProviderMessage returns the bounded provider error message.
+func (e *GenerationError) ProviderMessage() string {
+	if e == nil {
+		return ""
+	}
+	return e.providerMessage
+}
+
+// Category returns Generation for every generation failure.
+func (e *GenerationError) Category() ErrorCategory { return Generation }
+
+// Error intentionally excludes provider details from ordinary error text.
+func (e *GenerationError) Error() string {
+	if e == nil {
+		return ""
+	}
+	message := NewError(Generation, "provider generation failed", nil).Error()
+	if e.statusCode == 0 {
+		return message
+	}
+	return fmt.Sprintf("%s (HTTP %d)", message, e.statusCode)
+}
+
+// GoString keeps %#v formatting as safe as ordinary error formatting.
+func (e *GenerationError) GoString() string { return e.Error() }
+
+// Unwrap preserves the project-owned classified error and its hidden cause.
+func (e *GenerationError) Unwrap() error {
+	if e == nil {
+		return nil
+	}
+	return e.err
+}
+
 // CapacityError adds the safe backend identity to a capacity failure.
 type CapacityError struct {
 	BackendID string

internal/promptexec/promptexec_test.go

@@ -3,6 +3,7 @@ package promptexec
 import (
 	"context"
 	"errors"
+	"fmt"
 	"strings"
 	"testing"
 	"unicode/utf8"
@@ -172,8 +173,8 @@ func TestBoundDiagnosticAndErrorText(t *testing.T) {
 	if bounded[0] != "a<>b" {
 		t.Fatalf("invalid UTF-8 diagnostic = %q, want replacement", bounded[0])
 	}
-	validation := NewValidation(ValidationFailed, "json_schema", "daily.schema.json", values)
-	if len(validation.Diagnostics) != maxValidationDiagnostics || validation.Diagnostics[0] != "a<>b" {
+	validation := NewValidation(ValidationFailed, "json_schema", "daily.schema.json", 2, values)
+	if validation.RepairAttempts != 2 || len(validation.Diagnostics) != maxValidationDiagnostics || validation.Diagnostics[0] != "a<>b" {
 		t.Fatalf("validation = %#v, want bounded diagnostics", validation)
 	}
 
@@ -187,8 +188,53 @@ func TestBoundDiagnosticAndErrorText(t *testing.T) {
 	}
 }
 
+func TestGenerationErrorKeepsProviderDetailsOutOfRoutineFormatting(t *testing.T) {
+	cause := errors.New("provider response must stay hidden")
+	error := NewGenerationError(
+		429,
+		string([]byte{'c', 0xff})+strings.Repeat("界", 300),
+		strings.Repeat("type", 100),
+		string([]byte{'m', 0xff})+strings.Repeat("界", 4_200),
+		cause,
+	)
+	if error.StatusCode() != 429 || error.Category() != Generation || CategoryOf(error) != Generation {
+		t.Fatalf("generation error identity = %#v", error)
+	}
+	if utf8.RuneCountInString(error.ProviderCode()) != 256 || utf8.RuneCountInString(error.ProviderType()) != 256 || utf8.RuneCountInString(error.ProviderMessage()) != 4096 {
+		t.Fatalf("provider detail bounds = %d/%d/%d", utf8.RuneCountInString(error.ProviderCode()), utf8.RuneCountInString(error.ProviderType()), utf8.RuneCountInString(error.ProviderMessage()))
+	}
+	if !utf8.ValidString(error.ProviderCode()) || !utf8.ValidString(error.ProviderType()) || !utf8.ValidString(error.ProviderMessage()) {
+		t.Fatalf("provider details are not valid UTF-8: %#v", error)
+	}
+	if !errors.Is(error, cause) {
+		t.Fatal("errors.Is() = false, want preserved dependency cause")
+	}
+	var owned *Error
+	if !errors.As(error, &owned) || owned.Category() != Generation {
+		t.Fatalf("errors.As(*Error) = %#v, want project-owned generation error", owned)
+	}
+	want := "generation: provider generation failed (HTTP 429)"
+	if error.Error() != want || fmt.Sprintf("%#v", error) != want {
+		t.Fatalf("ordinary formatting = %q / %#v, want %q", error.Error(), error, want)
+	}
+	for _, private := range []string{cause.Error(), error.ProviderCode(), error.ProviderType(), error.ProviderMessage()} {
+		if strings.Contains(error.Error(), private) || strings.Contains(fmt.Sprintf("%#v", error), private) {
+			t.Fatalf("generation error leaks provider detail %q", private)
+		}
+	}
+
+	statusOnly := NewGenerationError(503, "", "", "", nil)
+	if statusOnly.Error() != "generation: provider generation failed (HTTP 503)" {
+		t.Fatalf("status-only error = %q", statusOnly)
+	}
+	var nilError *GenerationError
+	if nilError.StatusCode() != 0 || nilError.ProviderCode() != "" || nilError.ProviderType() != "" || nilError.ProviderMessage() != "" || nilError.Category() != Generation || nilError.Error() != "" || nilError.GoString() != "" || nilError.Unwrap() != nil {
+		t.Fatalf("nil generation error = %#v", nilError)
+	}
+}
+
 func TestContractCopiesMutableValues(t *testing.T) {
-	preparation := Preparation{InputHashes: map[string]string{"data_package": "input-hash"}}
+	preparation := Preparation{InputHashes: map[string]string{"data_package": "input-hash"}, Output: OutputContract{RepairAttempts: 3}}
 	preparationDebug := &PreparationDebug{
 		RenderedMessages: []RenderedMessage{{Role: "user", Content: "rendered input"}},
 		StructuredSchema: []byte("schema body"),
@@ -196,7 +242,7 @@ func TestContractCopiesMutableValues(t *testing.T) {
 	}
 	execution := Execution{
 		InputHashes: map[string]string{"data_package": "input-hash"},
-		Validation:  Validation{Diagnostics: []string{"validation detail"}},
+		Validation:  Validation{RepairAttempts: 2, Diagnostics: []string{"validation detail"}},
 		RawOutput:   []byte("generated output"),
 		Debug: &ExecutionDebug{
 			RawOutput:             []byte("provider output"),
@@ -217,13 +263,13 @@ func TestContractCopiesMutableValues(t *testing.T) {
 	execution.Debug.RawOutput[0] = 'x'
 	execution.Debug.ValidationDiagnostics[0] = "changed"
 
-	if preparationCopy.InputHashes["data_package"] != "input-hash" {
+	if preparationCopy.InputHashes["data_package"] != "input-hash" || preparationCopy.Output.RepairAttempts != 3 {
 		t.Fatalf("preparation copy = %#v", preparationCopy)
 	}
 	if debugCopy.RenderedMessages[0].Content != "rendered input" || string(debugCopy.StructuredSchema) != "schema body" || string(debugCopy.ParametersJSON) != `{"temperature":0.2}` {
 		t.Fatalf("preparation debug copy = %#v", debugCopy)
 	}
-	if executionCopy.InputHashes["data_package"] != "input-hash" || executionCopy.Validation.Diagnostics[0] != "validation detail" || string(executionCopy.RawOutput) != "generated output" || string(executionCopy.Debug.RawOutput) != "provider output" || executionCopy.Debug.ValidationDiagnostics[0] != "detailed validation" {
+	if executionCopy.InputHashes["data_package"] != "input-hash" || executionCopy.Validation.RepairAttempts != 2 || executionCopy.Validation.Diagnostics[0] != "validation detail" || string(executionCopy.RawOutput) != "generated output" || string(executionCopy.Debug.RawOutput) != "provider output" || executionCopy.Debug.ValidationDiagnostics[0] != "detailed validation" {
 		t.Fatalf("execution copy = %#v", executionCopy)
 	}
 }

internal/report/daily_report.go

@@ -12,9 +12,10 @@ func dailyDefinition() Definition {
 		ID:                          Daily,
 		Name:                        "Daily Report",
 		PromptID:                    "weather.daily_generated_text",
-		PromptVersion:         "2.0.0",
+		PromptVersion:               "2.1.0",
 		TemplateID:                  "daily",
 		GeneratedTextSchemaID:       "daily",
+		GeneratedTextRepairAttempts: 1,
 		ArtifactGroup:               "daily",
 		OutputName:                  "daily.md",
 		DistributorPathTemplates: []string{

internal/report/definition.go

@@ -33,6 +33,7 @@ type Definition struct {
 	PromptVersion               string
 	TemplateID                  string
 	GeneratedTextSchemaID       string
+	GeneratedTextRepairAttempts int
 	ArtifactGroup               string
 	OutputName                  string
 	DistributorPathTemplates    []string

internal/report/hourly_report.go

@@ -14,9 +14,10 @@ func hourlyDefinition() Definition {
 		ID:                          Hourly,
 		Name:                        "Hourly Report",
 		PromptID:                    "weather.hourly_generated_text",
-		PromptVersion:         "2.0.0",
+		PromptVersion:               "2.1.0",
 		TemplateID:                  "hourly",
 		GeneratedTextSchemaID:       "hourly",
+		GeneratedTextRepairAttempts: 1,
 		ArtifactGroup:               "hourly",
 		OutputName:                  "hourly.md",
 		DistributorPathTemplates: []string{

internal/report/period_test.go

@@ -82,8 +82,8 @@ func TestRegistryContainsOnlyPromptBackedReports(t *testing.T) {
 	}
 
 	for _, definition := range definitions {
-		if definition.PromptVersion != "2.0.0" {
-			t.Fatalf("%s PromptVersion = %q, want 2.0.0", definition.ID, definition.PromptVersion)
+		if definition.PromptVersion != "2.1.0" {
+			t.Fatalf("%s PromptVersion = %q, want 2.1.0", definition.ID, definition.PromptVersion)
 		}
 		if definition.PromptID == "" {
 			t.Fatalf("%s PromptID is empty", definition.ID)
@@ -91,6 +91,9 @@ func TestRegistryContainsOnlyPromptBackedReports(t *testing.T) {
 		if definition.TemplateID == "" || definition.GeneratedTextSchemaID == "" {
 			t.Fatalf("%s template/schema = %q/%q, want both set", definition.ID, definition.TemplateID, definition.GeneratedTextSchemaID)
 		}
+		if definition.GeneratedTextRepairAttempts != 1 {
+			t.Fatalf("%s repair attempts = %d, want 1", definition.ID, definition.GeneratedTextRepairAttempts)
+		}
 	}
 
 	if _, err := registry.Lookup(ID("three_day")); err == nil {

internal/report/today_report.go

@@ -10,9 +10,10 @@ func todayDefinition() Definition {
 		ID:                          Today,
 		Name:                        "Today Report",
 		PromptID:                    "weather.today_generated_text",
-		PromptVersion:         "2.0.0",
+		PromptVersion:               "2.1.0",
 		TemplateID:                  "today",
 		GeneratedTextSchemaID:       "today",
+		GeneratedTextRepairAttempts: 1,
 		ArtifactGroup:               "today",
 		OutputName:                  "today.md",
 		DistributorPathTemplates: []string{

internal/report/tomorrow_report.go

@@ -10,9 +10,10 @@ func tomorrowDefinition() Definition {
 		ID:                          Tomorrow,
 		Name:                        "Tomorrow Report",
 		PromptID:                    "weather.tomorrow_generated_text",
-		PromptVersion:         "2.0.0",
+		PromptVersion:               "2.1.0",
 		TemplateID:                  "tomorrow",
 		GeneratedTextSchemaID:       "tomorrow",
+		GeneratedTextRepairAttempts: 1,
 		ArtifactGroup:               "tomorrow",
 		OutputName:                  "tomorrow.md",
 		DistributorPathTemplates: []string{