Fix AS-20 selection handback wording - #28
Conversation
Clarify the legacy Phase 1A category instruction so ambiguous items still resolve to a canonical category and carry internal reviewer metadata instead of deferring selection. Refs briancl2/build-meta-analysis#164 Closes #27
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1e57ca9a21
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| 8. Preserve original raw tags in an internal note field if captured, but FINAL displayed category must be one canonical category only. | ||
| 8. Preserve original raw tags only in an internal note field if captured: | ||
| - For output, write exactly one `Category` value from the canonical taxonomy above. | ||
| - For ambiguous or unmapped items, choose the closest canonical category and add internal note metadata `Reviewer Action`: `review_category_mapping` with a brief mapping note. This metadata is not rendered in the final markdown. |
There was a problem hiding this comment.
For ambiguous or unmapped items, this asks the agent to add Reviewer Action metadata but also says it is not rendered in the final markdown. This legacy Phase 1A prompt only writes the discoveries markdown file, and repo-wide search found no consumer or sidecar artifact for Reviewer Action/review_category_mapping, so the mapping note is lost exactly when a reviewer would need it to audit the forced category choice. Persist it somewhere durable, such as a non-customer-facing notes section or sidecar file, instead of keeping it purely internal.
Useful? React with 👍 / 👎.
1 participant
Summary
Closes #27.
Refs briancl2/build-meta-analysis#164.
Validation
python3 /Users/briancl/repos/repo-auditor/scripts/as_signature_scan.py AS-20 /Users/briancl/repos/CustomerNewsletter->fired=falsemake check-> all required suites pass, expected private-fixture skipsmake review-> no critical/high findings; deterministic medium/low wording fixes applied without a third review loopBoundaries