Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create and manage Claude Code skills following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns, file paths, content patterns), enforcement levels (block, suggest, warn), hook mechanisms (UserPromptSubmit, PreToolUse), session tracking, and the 500-line rule.
.claude/skills/dokhacgiakhoa-skill-developer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 56% | 0% |
Comprehensive guide for creating and managing skills in Claude Code with auto-activation system, following Anthropic's official best practices including the 500-line rule and progressive disclosure pattern.
Automatically activates when you mention:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 20,597 | 18,248 | -11% | 1 | 1 | 0% | 2,772 | 3,343 | +21% | 0 | 0 | — |
case-02 | fail→pass | 15,001 | 9,408 | -37% | 1 | 1 | 0% | 1,553 | 2,534 | +63% | 0 | 0 | — |
case-03 | fail→pass | 16,692 | 12,529 | -25% | 1 | 1 | 0% | 1,932 | 2,208 | +14% | 0 | 0 | — |
case-04 | pass→pass | 18,668 | 15,202 | -19% | 1 | 1 | 0% | 2,287 | 2,683 | +17% | 0 | 0 | — |
case-05 | fail→pass | 9,425 | 2,936 | -69% | 1 | 1 | 0% | 1,765 | 1,427 | -19% | 0 | 0 | — |
case-06 | pass→pass | 5,743 | 7,702 | +34% | 1 | 1 | 0% | 889 | 1,358 | +53% | 0 | 0 | — |
case-07 | fail→pass | 12,249 | 8,879 | -28% | 1 | 1 | 0% | 1,037 | 1,622 | +56% | 0 | 0 | — |
case-08 | pass→pass | 4,850 | 7,978 | +64% | 1 | 1 | 0% | 735 | 1,408 | +92% | 0 | 0 | — |
case-09 | pass→pass | 11,466 | 8,015 | -30% | 1 | 1 | 0% | 1,935 | 1,475 | -24% | 0 | 0 | — |
case-10 | fail→pass | 13,071 | 9,019 | -31% | 1 | 1 | 0% | 2,253 | 2,562 | +14% | 0 | 0 | — |
case-11 | pass→pass | 21,135 | 13,981 | -34% | 1 | 1 | 0% | 2,621 | 3,234 | +23% | 0 | 0 | — |
case-12 | pass→pass | 9,660 | 10,737 | +11% | 1 | 1 | 0% | 1,556 | 1,822 | +17% | 0 | 0 | — |
case-13 | pass→pass | 12,495 | 4,208 | -66% | 1 | 1 | 0% | 2,168 | 1,574 | -27% | 0 | 0 | — |
case-14 | pass→pass | 3,737 | 7,894 | +111% | 1 | 1 | 0% | 551 | 1,351 | +145% | 0 | 0 | — |
case-15 | fail→pass | 10,344 | 2,419 | -77% | 1 | 1 | 0% | 919 | 1,321 | +44% | 0 | 0 | — |
case-16 | fail→pass | 15,900 | 2,332 | -85% | 1 | 1 | 0% | 1,800 | 1,281 | -29% | 0 | 0 | — |
case-17 | fail→pass | 12,524 | 7,329 | -41% | 1 | 1 | 0% | 1,316 | 1,323 | +1% | 0 | 0 | — |
case-18 | fail→pass | 15,322 | 2,933 | -81% | 1 | 1 | 0% | 1,596 | 1,488 | -7% | 0 | 0 | — |
case-19 | fail→pass | 6,314 | 3,112 | -51% | 1 | 1 | 0% | 1,061 | 1,445 | +36% | 0 | 0 | — |
case-20 | pass→pass | 16,723 | 10,533 | -37% | 1 | 1 | 0% | 2,003 | 1,882 | -6% | 0 | 0 | — |
case-21 | fail→pass | 7,893 | 8,851 | +12% | 1 | 1 | 0% | 1,302 | 1,585 | +22% | 0 | 0 | — |
case-22 | pass→pass | 12,412 | 9,841 | -21% | 1 | 1 | 0% | 1,486 | 2,744 | +85% | 0 | 0 | — |
case-23 | pass→pass | 14,424 | 20,579 | +43% | 1 | 1 | 0% | 2,950 | 4,143 | +40% | 0 | 0 | — |
case-24 | pass→pass | 15,620 | 15,075 | -3% | 1 | 1 | 0% | 2,540 | 4,025 | +58% | 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. 24 cases were attempted. The headline lift of +50 percentage points is the difference between those two pass rates over the 24 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.