Scan tool-authored casefile statements in the language check

[samrusani]

Summary

• CI language check now parses casefile.json and scans tool-authored fields (finding statements, record notes, rationale, questions, narrative, enforcement notes, skipped reasons).
• Claim texts and media titles are not scanned. Identity exemptions still require sibling registry evidence; claim-quote exemptions do not apply to those fields.
• Corrupt casefile targets fail closed. Fixture memos and casefiles are unchanged.

Test plan

• Five gates green locally: 759 tests, ruff, mypy, language check (29 files)
• CI on this PR (pytest matrix, wheel install, pip-audit)

[greptile-apps]

Greptile Summary

The PR extends the CI language checker to parse casefiles and scan selected tool-authored prose while preserving narrow evidence-backed identity exemptions and excluding quoted claims and media inputs.

• Adds casefile discovery, JSON parsing, field extraction, and fail-closed handling for unreadable or malformed JSON.
• Adds tests for scanned fields, excluded quoted inputs, identity verification, and corrupt targets.
• Updates architecture, decisions, future-work, and handoff documentation to describe the expanded gate.

Confidence Score: 4/5

The PR appears safe to merge after the non-blocking cleanup of person-like names from the changed test descriptions.

The casefile scanner covers the intended generated-prose fields and preserves the documented exemptions; the only accepted concern is a repository-policy violation in test docstrings.

Files Needing Attention: tests/test_language.py

Important Files Changed

Filename	Overview
scripts/check_language.py	Adds casefile target discovery, selective extraction of tool-authored fields, evidence-backed identity exemptions, and malformed-JSON failure handling.
tests/test_language.py	Adds broad casefile-scan coverage, but changed test descriptions use person-like names contrary to the repository’s description/comment policy.
src/coldscreen/language.py	Updates shared language-policy documentation to distinguish memo claim exemptions from casefile field scanning.
DECISIONS.md	Records the selected casefile fields, exemption behavior, excluded quoted inputs, and failure policy.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Language-check target] --> B{casefile.json?}
  B -- No --> C[Scan memo or template text]
  B -- Yes --> D[Parse JSON]
  D --> E{Readable and valid JSON?}
  E -- No --> F[Fail language check]
  E -- Yes --> G[Extract selected tool-authored fields]
  G --> H[Verify identity exemptions against registry evidence]
  H --> I[Scan each field for banned terms]
  C --> J{Banned term found?}
  I --> J
  J -- Yes --> F
  J -- No --> K[Pass target]

Loading

Prompt To Fix All With AI

### Issue 1
tests/test_language.py:589-590
**Remove names from test descriptions**

The changed docstrings include person-like full names despite the repository rule prohibiting personally identifiable information in descriptions and comments. Replace these references, including the adjacent identity-test description, with explicitly synthetic placeholders.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

_{Reviews (1): Last reviewed commit: "Scan tool-authored casefile statements i..." | Re-trigger Greptile}

[greptile-apps]

Remove names from test descriptions

The changed docstrings include person-like full names despite the repository rule prohibiting personally identifiable information in descriptions and comments. Replace these references, including the adjacent identity-test description, with explicitly synthetic placeholders.

Rule Used: Make sure no personally identifiable information i... (source)

Prompt To Fix With AI

This is a comment left during a code review.
Path: tests/test_language.py
Line: 589-590

Comment:
**Remove names from test descriptions**

The changed docstrings include person-like full names despite the repository rule prohibiting personally identifiable information in descriptions and comments. Replace these references, including the adjacent identity-test description, with explicitly synthetic placeholders.

**Rule Used:** Make sure no personally identifiable information i... ([source](https://app.greptile.com/900-labs/-/custom-context?memory=930389e5-d152-4d6b-9814-6e9c4f9b209c))

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!