[mlieberman85]

Stacked on #670. Review the last 3 commits: the spec change, the procedures, and the skipped list.

Summary

darnit-comply told the agent to "follow the /darnit-data skill" and "follow the /darnit-audit skill". Nothing loads another skill's text, so the comply run improvised data collection from one line and asked questions differently from a darnit-data run (#654).

• darnit.server.procedures holds the steps once:

  • get_pending_data returns them as procedure;
  • get_judgment_requests returns the judging steps on every page;
  • the baseline confirm_project_data result ends with the next get_pending_data call.

  Calls name only the owner and repo the caller passed, never ones read from the checkout's remote, and hold no candidate value.

• skipped: a get_judgment_requests page names the controls it left out because they already have a stored PASS candidate, and the procedure says not to request them again. In the audit_openssf_baseline JSON output is about 1 MB and exceeds the MCP client's result limit #644 e2e runs the agent asked for them again and got empty pages.

• Skills:

  • The skills defer to the procedures, and a test checks that no skill tells the agent to follow another one.
  • Comply and remediate pass owner and repo to remediate_audit_findings.

• Spec: 5.4 and 7.x.

Refs #654.

Type of Change

• New feature (non-breaking change adding functionality)

Framework Changes Checklist

• Updated framework spec (docs/architecture/framework-design.md)
• Ran uv run python scripts/validate_sync.py --verbose and it passes

Testing

• uv run pytest tests/ on the whole rebased stack (18bd63e): 5286 passed, 18 skipped
• e2e: see the table below
• uv run ruff check .

End-to-end

darnit-e2e: headless /darnit:darnit-audit openssf-baseline through Claude Code. Confirm tools are disallowed and the repository is named in the prompt.

darnit	darnit repo	wtf-frontend	cosign-installer
main (2026-10-20; cosign-installer on 603bdda)	387s, 44 turns	415s, 47 turns	297s
#653 only (2026-10-21)	299s, 57 turns	250s, 55 turns	185s, 39 turns
full stack, rebased on 8070949 (2026-10-23)	282s, 50 turns, 30 judgments	145s, 23 turns, 15 judgments	141s, 30 turns

• Results: per-control results match main. The only differences are controls the model judged, and the effect of operator-side answers on wtf-frontend, which needs the repository named. wtf-frontend also shows QA-03.01 and QA-07.01 passing because their platform settings changed on GitHub.
• Stored candidates skew the comparison. A control with a stored PASS candidate for unchanged evidence is skipped, so run-to-run times depend on how many are stored. darnit's HEAD moved to 8070949, which invalidated its stored candidates, so it was judged in full (30 controls).
• Denials:
  • darnit: one page of get_judgment_requests was larger than Claude Code's tool-result limit and was saved to a file. The agent's attempt to read it with Bash was denied. This is the oversized single-request follow-up.
  • wtf-frontend: the agent tried to run zizmor itself, and that was denied.

AI assistance

• No AI assistance was used
• AI assistance was used

Claude Code (Claude Opus) wrote the change, tests and this description; I reviewed them and ran the e2e.

---

Drafted with AI assistance (Claude), reviewed and filed by me. Measurements quoted were executed and verified.

[Marc-cn]

Reviewed its own three commits. One source for the procedure text, calls name only the repository and checkout the caller passed (placeholders otherwise, never a detected value), confirm_project_data ends with the next call, skipped lists controls with a stored candidate. One leftover from #670: a PENDING llm_judgment control with no llm_consultation is still absent from requests, remaining and skipped, so an agent cannot tell why it never shows up. Rare, but listing it (with why) would close the gap. Good to merge in order.