Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when auditing a large local skill collection, identifying duplicate or imported skills, comparing skill roots, or deciding what to keep, disable, or archive across Codex and adjacent agent skill directories.
.claude/skills/cnfjlhj-skills-governance/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -47% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-22 | ✗→✓ | ▲ Improved | -42% | 0% |
Use the bundled scanner as the source of truth for local skill inventory. Base recommendations on real directories and SKILL.md frontmatter instead of stale catalogs or memory.
Invoke the bundled scan.js from this skill directory.
node scan.js --mode codex --format markdownnode scan.js --mode all --format markdownnode scan.js --mode all --duplicates-only --format markdownnode scan.js --mode all --format jsoncodex mode first.all mode.[[skills.config]] disable suggestions over editing individual SKILL.md files unless the user explicitly asks for direct file edits.Report at least:
plugin-import, system, or backup-likeA good run:
name| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 16,244 | 6,178 | -62% | 1 | 1 | 0% | 3,021 | 745 | -75% | 0 | 0 | — |
case-02 | fail→fail | 19,697 | 6,095 | -69% | 1 | 1 | 0% | 3,356 | 699 | -79% | 0 | 0 | — |
case-03 | fail→fail | 18,844 | 5,479 | -71% | 1 | 1 | 0% | 3,293 | 807 | -75% | 0 | 0 | — |
case-04 | fail→pass | 9,583 | 5,466 | -43% | 1 | 1 | 0% | 1,553 | 1,382 | -11% | 0 | 0 | — |
case-05 | pass→pass | 6,038 | 2,839 | -53% | 1 | 1 | 0% | 927 | 884 | -5% | 0 | 0 | — |
case-06 | pass→pass | 7,505 | 7,755 | +3% | 1 | 1 | 0% | 1,172 | 1,737 | +48% | 0 | 0 | — |
case-07 | fail→pass | 27,488 | 4,762 | -83% | 1 | 1 | 0% | 2,384 | 1,275 | -47% | 0 | 0 | — |
case-08 | fail→pass | 7,828 | 3,135 | -60% | 1 | 1 | 0% | 1,151 | 963 | -16% | 0 | 0 | — |
case-09 | pass→pass | 9,801 | 2,238 | -77% | 1 | 1 | 0% | 1,416 | 814 | -43% | 0 | 0 | — |
case-10 | pass→pass | 14,849 | 2,988 | -80% | 1 | 1 | 0% | 2,541 | 915 | -64% | 0 | 0 | — |
case-11 | fail→pass | 7,740 | 3,145 | -59% | 1 | 1 | 0% | 1,171 | 931 | -20% | 0 | 0 | — |
case-12 | pass→pass | 16,826 | 3,195 | -81% | 1 | 1 | 0% | 1,314 | 948 | -28% | 0 | 0 | — |
case-13 | fail→fail | 8,033 | 7,114 | -11% | 1 | 1 | 0% | 1,308 | 780 | -40% | 0 | 0 | — |
case-22 | fail→pass | 8,926 | 2,932 | -67% | 1 | 1 | 0% | 1,531 | 883 | -42% | 0 | 0 | — |
case-14 | pass→pass | 7,972 | 6,359 | -20% | 1 | 1 | 0% | 1,181 | 1,326 | +12% | 0 | 0 | — |
case-15 | fail→fail | 4,060 | 9,289 | +129% | 1 | 1 | 0% | 534 | 1,801 | +237% | 0 | 0 | — |
case-16 | fail→fail | 8,971 | 3,411 | -62% | 1 | 1 | 0% | 1,521 | 998 | -34% | 0 | 0 | — |
case-17 | pass→pass | 15,268 | 8,601 | -44% | 1 | 1 | 0% | 2,396 | 1,691 | -29% | 0 | 0 | — |
case-18 | pass→pass | 12,231 | 6,329 | -48% | 1 | 1 | 0% | 1,950 | 1,386 | -29% | 0 | 0 | — |
case-19 | pass→pass | 11,509 | 4,283 | -63% | 1 | 1 | 0% | 1,805 | 1,071 | -41% | 0 | 0 | — |
case-20 | fail→pass | 15,590 | 9,162 | -41% | 1 | 1 | 0% | 2,640 | 2,022 | -23% | 0 | 0 | — |
case-21 | pass→pass | 20,590 | 2,253 | -89% | 1 | 1 | 0% | 1,364 | 755 | -45% | 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 18 counted toward the lift figure. The other 4 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 +27 percentage points is the difference between those two pass rates over the 18 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.