Align prompts and strengthen validation context

This commit is contained in:
2026-09-25 12:23:10 +00:00
parent 3eb73e70fc
commit 325c4ea7d3
23 changed files with 329 additions and 133 deletions

View File

@@ -1,6 +1,6 @@
# Audita Prompt Registry
`internal/prompts` embeds nine PromptKit YAML definitions under
`internal/prompts` embeds seven PromptKit YAML definitions under
`assets/prompts`. The IDs and selected version remain:
- `modules.glossary.proposal` (`v1`)
@@ -10,17 +10,15 @@
- `validators.spoken_form_plausibility` (`v1`)
- `validators.meaning_reversal_review` (`v1`)
- `validators.editorial_review` (`v1`)
- `validators.grammar_review` (`v1`)
- `validators.spoken_word_review` (`v1`)
PromptKit reads the embedded definitions and renders their system and user
messages from named string variables. The role-bound adapter will supply
messages from named string variables. The role-bound adapter supplies
`HardeningPolicy` from the single embedded `assets/shared/prompt_hardening.md`
fragment after caller variables, so callers cannot replace that policy. Proposal prompts receive the
section JSON, glossary JSON where used, and transcript description block.
Validator prompts receive the correction payload and description block.
The prompt assets retain the prior task instructions and output-schema
association. The role-bound runtime will supply the complete JSON Schema output contract
association. The role-bound runtime supplies the complete JSON Schema output contract
and repair count; the definition does not own an operator-facing repair default.
The correction-set schema requires a string replacement and accepts an empty
string. Proposal guidance uses it for eligible span deletion and permits literal
@@ -41,3 +39,13 @@ response-schema key, and logical-call attribution. Module and validator callers
pass these values through the role-bound adapter. See the
[LLM runtime](llm-runtime.md) for runtime assembly and
[configuration](../config.md) for operator-selected profiles.
Proposal instructions describe literal matching, each module's replacement policy,
source-based proposal identity, and confidence in whether an edit is warranted.
Validator instructions distinguish correction indices from segment IDs and confidence
in the verdict from the verdict itself. The grammar article exception and glossary
mishearing exception apply consistently across their instructions.
Only the three active validator definitions are retained. The unused internal
`grammar_review` and `spoken_word_review` definitions and type selections were
removed; built-in chains and operator configuration are unchanged.