Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Validate maintainer docs structure, standards freshness, manifest paths, and plan state. Run this after modifying any file under .agents/docs/standards/, .agents/docs/plans/, .agents/docs/architecture/, or AGENTS.md — it catches stale metadata, broken manifest paths, and plan state drift before CI does.
.claude/skills/liaohch3-legibility-check/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 205% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -47% | 0% |
Run deterministic checks for maintainer docs that mirror what CI enforces via .github/workflows/legibility.yml. Catching these locally saves a round-trip to CI.
.agents/docs/standards/*.md must have frontmatter with owner, last_reviewed (ISO date), and source_of_truth. Files reviewed more than 60 days ago produce a warning..agents/docs/architecture/manifest.yaml under expected_paths: must exist in the repo..agents/docs/plans/**/*.md must have a status frontmatter field (active, completed, or cancelled). Completed plans must not contain unchecked - [ ] checkboxes (outside fenced code blocks).bashuv run python scripts/check_legibility.py
Options:
--freshness-days N — change the staleness threshold (default: 60)--strict-freshness — promote stale warnings to failures--repo-root PATH — override repo root (default: cwd)| Failure | Fix | |---------|-----| | missing frontmatter key 'X' | Add the missing key to the YAML frontmatter block at the top of the file | | last_reviewed must be ISO date | Use YYYY-MM-DD format | | last_reviewed ... is stale | Update last_reviewed to today's date after reviewing the content | | expected path missing: X | Either create the file or remove the stale entry from manifest.yaml | | status must be one of [...] | Add status: active (or completed/cancelled) to plan frontmatter | | completed plan still contains unchecked TODO | Check off remaining items or change status back to active |
Re-run the check to confirm all issues are resolved before committing:
bashuv run python scripts/check_legibility.py && echo "All clear"
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,743 | 1,497 | -60% | 1 | 1 | 0% | 191 | 645 | +238% | 0 | 0 | — |
case-02 | fail→fail | 5,655 | 1,531 | -73% | 1 | 1 | 0% | 896 | 657 | -27% | 0 | 0 | — |
case-03 | fail→pass | 4,463 | 3,267 | -27% | 1 | 1 | 0% | 226 | 689 | +205% | 0 | 0 | — |
case-04 | fail→pass | 4,014 | 1,546 | -61% | 1 | 1 | 0% | 664 | 781 | +18% | 0 | 0 | — |
case-05 | fail→pass | 5,977 | 1,665 | -72% | 1 | 1 | 0% | 1,136 | 778 | -32% | 0 | 0 | — |
case-06 | fail→pass | 7,357 | 2,535 | -66% | 1 | 1 | 0% | 1,471 | 919 | -38% | 0 | 0 | — |
case-07 | fail→pass | 8,867 | 2,320 | -74% | 1 | 1 | 0% | 1,724 | 913 | -47% | 0 | 0 | — |
case-08 | pass→pass | 6,466 | 2,353 | -64% | 1 | 1 | 0% | 1,282 | 927 | -28% | 0 | 0 | — |
case-09 | fail→pass | 8,553 | 1,284 | -85% | 1 | 1 | 0% | 1,600 | 687 | -57% | 0 | 0 | — |
case-10 | fail→pass | 5,744 | 1,524 | -73% | 1 | 1 | 0% | 968 | 663 | -32% | 0 | 0 | — |
case-11 | fail→pass | 13,247 | 1,673 | -87% | 1 | 1 | 0% | 2,352 | 725 | -69% | 0 | 0 | — |
case-12 | pass→pass | 4,875 | 1,317 | -73% | 1 | 1 | 0% | 886 | 736 | -17% | 0 | 0 | — |
case-13 | fail→pass | 9,898 | 1,293 | -87% | 1 | 1 | 0% | 1,859 | 695 | -63% | 0 | 0 | — |
case-14 | fail→pass | 11,500 | 3,135 | -73% | 1 | 1 | 0% | 1,955 | 1,010 | -48% | 0 | 0 | — |
case-15 | pass→pass | 5,011 | 3,220 | -36% | 1 | 1 | 0% | 992 | 1,107 | +12% | 0 | 0 | — |
case-16 | fail→pass | 3,415 | 2,057 | -40% | 1 | 1 | 0% | 608 | 853 | +40% | 0 | 0 | — |
case-17 | pass→pass | 7,228 | 2,661 | -63% | 1 | 1 | 0% | 1,316 | 952 | -28% | 0 | 0 | — |
case-18 | fail→pass | 4,548 | 1,792 | -61% | 1 | 1 | 0% | 896 | 818 | -9% | 0 | 0 | — |
case-19 | pass→pass | 9,982 | 1,660 | -83% | 1 | 1 | 0% | 1,659 | 732 | -56% | 0 | 0 | — |
case-20 | pass→pass | 6,678 | 3,441 | -48% | 1 | 1 | 0% | 1,313 | 1,187 | -10% | 0 | 0 | — |
case-21 | pass→pass | 8,420 | 5,608 | -33% | 1 | 1 | 0% | 1,730 | 1,649 | -5% | 0 | 0 | — |
case-22 | pass→pass | 6,617 | 5,332 | -19% | 1 | 1 | 0% | 1,380 | 1,591 | +15% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +55 percentage points is the difference between those two pass rates over the 20 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.