v0.6.4: skill audit — author's-side staleness read (Stream 6 follow-on)

[thrillmot]

Summary

Closes the "author's-side" arrow of the SkDD loop. logmind skill audit walks every .claude/skills/*/SKILL.md and reports staleness signals:

• name + path + bytes
• last touched — last git commit touching SKILL.md (falls back to file mtime if not in git)
• decision count — # times the skill name appears in docs/decisions.md + docs/decisions-branches/*.md
• status: active (everyday) / aging (last touched > 90d) / ghost (decision_count==0 AND bytes>2000 → loaded but never iterated)

Pairs with clud-bug's usage --health (v0.6.28+):

• audit: what's HERE, how stale (author + filesystem side)
• usage --health: what's USED, how often cited (load + review side)

Surface

• logmind skill audit — human-readable table with summary line.
• logmind skill audit --json — machine-readable array.

What changed

• src/logmind/core/skill_cli.py — new audit_skills, classify_audit_row, _git_last_touched.
• src/logmind/cli.py — new @skill.command("audit") wiring.
• tests/test_skill_cli.py — 13 new tests (directory walking, decision counting from main + branch files, every status bucket, CLI rendering, JSON output, edge cases).

Test plan

• pytest -q — 772 tests pass (758 prior + 14 new)
• CLI smoke: scaffold a SKILL.md + log a decision mentioning it, run logmind skill audit, verify rendered table + status classification
• Self-review on this PR via clud-bug-review

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
logmind-site	Ready	Preview, Comment	Jun 1, 2026 4:42am

[github-actions]

🐛 Q7-logmind bench

ok: 4-angle Q7-logmind compliance
  per-call       -10% bytes vs git equivalent      ✅ saver
  worst-case     -58% even on never-read           ✅ saver
  per-session    (stub — not yet implemented)
  org-cumulative (stub — not yet implemented)

Updated 2026-06-01T04:42Z. See workflow run for full JSON.

[claude]

Claude finished @thrillmot's task in 3m 31s —— View job

---

🐛 Clud Bug review

This round: 0 critical · 0 minor · 0 resolved from prior · 0 still open

Found: 0 🔴 / 0 🟡 / 0 🟣

Per-skill scan

• [critical-issues-only]: scanned all paths. 0 findings.
• [evidence-based-review]: applied to all findings. ✓ all anchored.
• [respect-existing-conventions]: deferred import datetime as _dt inside functions matches existing bench/log conventions. 0 findings.

Tasks

• Load tools
• Read skills
• Fetch PR diff
• Review src/logmind/core/skill_cli.py
• Review src/logmind/cli.py
• Review tests/test_skill_cli.py
• Check remaining files
• Post inline findings (none)
• Emit structured summary

No critical issues found. The implementation is correct: imports are clean (Optional, subprocess, default_skills_dir all present), the git subprocess is invocation-safe (list args + -- separator), classify_audit_row priority (ghost → aging → active) matches spec, mtime fallback for non-git files works, and test coverage hits all status buckets + CLI rendering + JSON output + edge cases.

Skills referenced: [critical-issues-only, evidence-based-review, respect-existing-conventions]

[github-actions]

🐛 Clud Bug review

This round: 0 critical · 0 minor · 0 resolved from prior · 0 still open

Found: 0 🔴 / 0 🟡 / 0 🟣

Per-skill scan

• [critical-issues-only]: scanned all paths. 0 findings.
• [evidence-based-review]: applied to all findings. 0 findings.
• [respect-existing-conventions]: checked deferred import pattern against existing conventions. 0 findings.

Skills referenced: [critical-issues-only, evidence-based-review, respect-existing-conventions]