Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Audit and quality-gate scripts for the template research framework. Covers documentation linting, filepath audits, mock-usage checking, template drift, and confidentiality / git-guard checks.
.claude/skills/docxology-template-audit/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -59% | 0% |
Audit and quality-gate scripts in scripts/audit/.
Load this skill when you need to:
| Script | Exit 0 means | |--------|-------------| | lint_docs.py | All doc lints pass | | check_tracked_projects.py | No private projects staged/committed | | check_template_drift.py | Exemplars are up-to-date | | verify_no_mocks.py | No prohibited mock-framework imports/calls detected |
All scripts use parents[2] from scripts/audit/ — three levels to repo root.
check_tracked_all.py runs in pre-push and composes every resource-pool guard;keep each constituent check fast.
audit_documentation.py is advisory; non-zero exit is a warning, not a gate.copy_exemplar.py modifies tracked files — review diff before committing.verify_no_mocks.py --inventory --max-dependency-replacements 0 is blockingin CI. The default lexical pass and the semantic zero-debt gate are distinct contracts; environment isolation remains separately classified.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,438 | 7,241 | -3% | 1 | 1 | 0% | 1,476 | 646 | -56% | 0 | 0 | — |
case-02 | fail→fail | 6,431 | 2,639 | -59% | 1 | 1 | 0% | 1,119 | 693 | -38% | 0 | 0 | — |
case-03 | fail→fail | 3,212 | 4,708 | +47% | 1 | 1 | 0% | 309 | 593 | +92% | 0 | 0 | — |
case-04 | pass→pass | 6,450 | 3,913 | -39% | 1 | 1 | 0% | 1,327 | 1,105 | -17% | 0 | 0 | — |
case-05 | pass→pass | 9,292 | 5,795 | -38% | 1 | 1 | 0% | 1,775 | 1,474 | -17% | 0 | 0 | — |
case-06 | pass→pass | 7,522 | 6,306 | -16% | 1 | 1 | 0% | 1,409 | 1,666 | +18% | 0 | 0 | — |
case-07 | fail→pass | 12,364 | 5,058 | -59% | 1 | 1 | 0% | 2,176 | 1,153 | -47% | 0 | 0 | — |
case-08 | fail→pass | 8,991 | 2,823 | -69% | 1 | 1 | 0% | 1,579 | 719 | -54% | 0 | 0 | — |
case-09 | fail→pass | 4,314 | 2,753 | -36% | 1 | 1 | 0% | 921 | 874 | -5% | 0 | 0 | — |
case-10 | fail→pass | 7,146 | 4,256 | -40% | 1 | 1 | 0% | 1,166 | 954 | -18% | 0 | 0 | — |
case-11 | fail→pass | 11,475 | 2,287 | -80% | 1 | 1 | 0% | 1,739 | 706 | -59% | 0 | 0 | — |
case-12 | pass→pass | 5,676 | 2,203 | -61% | 1 | 1 | 0% | 997 | 706 | -29% | 0 | 0 | — |
case-13 | pass→pass | 8,846 | 4,892 | -45% | 1 | 1 | 0% | 1,575 | 1,123 | -29% | 0 | 0 | — |
case-14 | pass→pass | 11,089 | 4,524 | -59% | 1 | 1 | 0% | 1,834 | 766 | -58% | 0 | 0 | — |
case-15 | pass→pass | 7,249 | 3,265 | -55% | 1 | 1 | 0% | 1,268 | 880 | -31% | 0 | 0 | — |
case-16 | fail→pass | 9,927 | 4,226 | -57% | 1 | 1 | 0% | 1,723 | 1,151 | -33% | 0 | 0 | — |
case-17 | fail→pass | 8,871 | 2,675 | -70% | 1 | 1 | 0% | 1,684 | 613 | -64% | 0 | 0 | — |
case-18 | pass→pass | 16,443 | 2,557 | -84% | 1 | 1 | 0% | 1,493 | 799 | -46% | 0 | 0 | — |
case-19 | fail→pass | 16,281 | 2,022 | -88% | 1 | 1 | 0% | 2,394 | 715 | -70% | 0 | 0 | — |
case-20 | pass→pass | 2,977 | 2,644 | -11% | 1 | 1 | 0% | 479 | 905 | +89% | 0 | 0 | — |
case-21 | pass→pass | 6,642 | 2,193 | -67% | 1 | 1 | 0% | 1,071 | 740 | -31% | 0 | 0 | — |
case-22 | pass→pass | 12,561 | 9,872 | -21% | 1 | 1 | 0% | 1,915 | 2,041 | +7% | 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 +36 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.