Add bounded sources and strict YAML parsing

docs/roadmap/implementation.md

@@ -2,7 +2,7 @@
 
 ## Status and ownership
 
-Status: implementation in progress; Stages 1–2 are complete and later stages are
+Status: implementation in progress; Stages 1–3 are complete and later stages are
 pending.
 
 Audience: a `gpt-5.6-terra` coding agent implementing one stage per prompt, in
@@ -383,7 +383,30 @@ Acceptance:
 
 ## Stage 3 — Extract source reading and enforce YAML policy
 
-Status: pending.
+Status: complete.
+
+Completion evidence (2026-09-06):
+
+- Added opaque source constructors and bounded file/byte/reader acquisition in
+  [`source.go`](../../source.go), including byte snapshots, regular-file checks,
+  deterministic I/O codes, caller-owned reader lifetime, and one atomic claim
+  shared by source copies. `NoSource` creates no source record.
+- Pinned `go.yaml.in/yaml/v4 v4.0.0-rc.6` solely behind `internal/yaml` and added
+  the parser-independent ordered tree in `internal/document`. The adapter uses the
+  streaming loader, checks for a second document explicitly, applies a calibrated
+  private parser depth plugin, enforces owned depth/node limits, preserves numeric
+  spelling and locations, and returns only value-free internal issues.
+- Generalized the strict-source/document cases recorded in the [extraction
+  inventory](../internal/extraction.md), created the [current configuration input
+  contract](../config.md), and added an AST import-boundary regression. No typed
+  decoder, composition, filesystem graph, or public `Load` was introduced.
+- Tests cover real file failures and reopening, byte ownership, fragmented/error
+  readers, exact/overflow byte limits, concurrent copied-source claims, all YAML
+  policy categories, trailing/empty documents, duplicate locations, lexical wide
+  numbers, block/flow depth calibration, node limits, safe parser errors, and
+  adapter isolation. Verified `go test ./...`, `go vet ./...`, `go build ./...`,
+  `go mod tidy -diff`, and `git diff --check` with Go 1.26.7 on linux/amd64. No
+  deviations were required; typed binding and decoding remain later work.
 
 Read: feature [YAML policy](initial.md#yaml-and-scalar-policy),
 [source constructors](initial.md#entry-point-and-reusable-application-definition).

docs/roadmap/initial.md

@@ -423,31 +423,10 @@ configuration is never returned as a successful, usable result.
 
 ## YAML and scalar policy
 
-The supported model is mappings with string keys, sequences, scalars, and null,
-with YAML comments and block/flow syntax. Require exactly one document with a
-mapping root. Reject empty input, additional documents (including empty ones),
-duplicate mapping keys, explicit tags, anchors, aliases, and merge keys (`<<`).
-Mapping keys must resolve to strings under this scalar policy; quote keys
-that would otherwise be booleans, numbers, or null. Compare decoded string keys
-without case folding or Unicode normalization. A quoted `"<<"` is also reserved and rejected.
-
-Use a deliberately limited scalar vocabulary independent of parser defaults:
-
-- Quoted and block scalars are strings.
-- Plain `true` and `false` are booleans; plain `null`, `~`, and an empty value
-  are null. Other case variants of these reserved literals are rejected when used as
-  plain values.
-- Numbers use JSON decimal syntax: optional minus, integer part with no leading
-  zero except zero itself, optional fraction, and optional decimal exponent.
-- Reject YAML numeric extensions such as hexadecimal, octal, digit separators,
-  leading plus, and non-finite values. Reject legacy boolean words such as
-  `yes`, `no`, `on`, and `off` unless quoted (case-insensitively).
-- Date/time-looking plain text remains a string, never an implicit timestamp.
-  Other ordinary plain text is a string.
-
-Unsupported scalar forms must produce a policy error rather than inherit a
-parser conversion. Preserve numeric lexical information until decoding so
-large integers are not rounded through an intermediate floating-point value.
+Source ownership, the supported YAML subset, scalar grammar, fixed input limits,
+and safe failure behavior are implemented. The [current configuration input
+contract](../config.md) now owns those definitions. Typed scalar conversion into
+application values remains target work in [typed decoding](#typed-decoding).
 
 ## Paths and presence
 
@@ -770,11 +749,6 @@ type FilesystemContext struct {
     BaseDir string
     Root    string
 }
-
-func File(path string) Source
-func Bytes(label string, data []byte) Source
-func Reader(label string, input io.Reader) Source
-func NoSource() Source
 ```
 
 `Load` accepts zero or one options value; multiple values are an options error,
@@ -792,11 +766,11 @@ result. Callbacks must not retain these contexts or configuration pointers; shar
 closures must be safe for concurrent calls if the application loads concurrently.
 ConfigKit does not synchronize application-owned captured state.
 
-`Source` is opaque with the constructors above, not an arbitrary provider
-interface. `Bytes` snapshots its input; `Reader` is consumed during the load and
-is not closed by ConfigKit. A reader source is single-use. File sources are read
-on each load. Constructor argument errors surface as load diagnostics. Applications must use
-non-secret source labels because labels appear in diagnostics and source reports.
+`Source` and its constructors are implemented; [`source.go`](../../source.go) and
+the [configuration input contract](../config.md#sources) own their current behavior.
+The loading and filesystem-option behavior below remains target work.
+Applications must use non-secret source labels because labels appear in diagnostics
+and source reports.
 Filesystem context follows the [source rules](#explicit-imports-and-source-identity):
 a file's own directory supplies its base, and an explicit context can set its
 root. Supplying `BaseDir` for a file is an options error; bytes/readers require