Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create comprehensive API documentation with examples, authentication guides, and SDKs. Use when creating comprehensive API documentation. Trigger with phrases like "generate API docs", "create API documentation", or "document the API".
.claude/skills/jeremylongshore-generating-api-docs/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 52% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-22 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 7% | 0% |
Create comprehensive, interactive API documentation from OpenAPI specifications with runnable code examples, authentication guides, error reference tables, and SDK quick-start tutorials. Generate documentation sites using Redoc, Stoplight Elements, or Swagger UI with custom branding, versioned navigation, and full-text search.
summary, description, parameter descriptions, and at least one example per request/response.info.description or x-documentation extensions.See ${CLAUDE_SKILL_DIR}/references/implementation.md for the full implementation guide.
${CLAUDE_SKILL_DIR}/docs/site/ - Generated documentation website (HTML/CSS/JS)${CLAUDE_SKILL_DIR}/docs/guides/authentication.md - Authentication flow guide with code examples${CLAUDE_SKILL_DIR}/docs/guides/getting-started.md - Quick-start tutorial for first API call${CLAUDE_SKILL_DIR}/docs/reference/errors.md - Complete error code reference with resolution steps${CLAUDE_SKILL_DIR}/docs/examples/ - Per-endpoint code examples in multiple languages${CLAUDE_SKILL_DIR}/docs/config/redoc.yaml - Documentation generator configuration with branding| Error | Cause | Solution | |-------|-------|----------| | Missing examples | OpenAPI spec lacks example or examples for request/response schemas | Generate examples from schema using Faker-based data; flag missing examples in spec linting | | Stale documentation | Docs deployed from an older spec version than the running API | Tie doc deployment to API deployment pipeline; version docs alongside API releases | | Broken Try-It requests | CORS not configured for documentation domain on the API server | Add documentation domain to CORS Access-Control-Allow-Origin; or use a proxy for Try-It requests | | Code example errors | Generated code example uses deprecated SDK method or wrong import path | Auto-test code examples against a staging API; version examples alongside SDK releases | | Search not working | Full-text search index not rebuilt after content update | Include search index regeneration in documentation build pipeline; verify Algolia/Lunr config |
Refer to ${CLAUDE_SKILL_DIR}/references/errors.md for comprehensive error patterns.
Stripe-style API docs: Generate a three-column documentation layout with navigation on the left, description in the center, and code examples on the right, with language switcher for curl/Node/Python/Ruby.
Versioned documentation site: Host v1 and v2 documentation side-by-side with a version switcher dropdown, and a changelog page highlighting breaking changes and migration steps between versions.
Developer portal: Combine API reference docs with getting-started tutorials, use-case guides, webhooks documentation, and SDK installation instructions in a single searchable portal.
See ${CLAUDE_SKILL_DIR}/references/examples.md for additional examples.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 42,176 | 15,927 | -62% | 1 | 1 | 0% | 8,281 | 1,665 | -80% | 0 | 0 | — |
case-02 | fail→fail | 42,375 | 7,758 | -82% | 1 | 1 | 0% | 8,275 | 1,264 | -85% | 0 | 0 | — |
case-03 | fail→fail | 36,853 | 40,247 | +9% | 1 | 1 | 0% | 8,285 | 1,577 | -81% | 0 | 0 | — |
case-04 | pass→pass | 27,623 | 19,655 | -29% | 1 | 1 | 0% | 4,974 | 5,312 | +7% | 0 | 0 | — |
case-05 | pass→pass | 11,938 | 14,259 | +19% | 1 | 1 | 0% | 2,189 | 3,897 | +78% | 0 | 0 | — |
case-06 | pass→pass | 20,691 | 36,658 | +77% | 1 | 1 | 0% | 4,725 | 9,314 | +97% | 0 | 0 | — |
case-07 | pass→pass | 24,452 | 17,585 | -28% | 1 | 1 | 0% | 3,575 | 4,307 | +20% | 0 | 0 | — |
case-08 | fail→fail | 13,193 | 12,873 | -2% | 1 | 1 | 0% | 2,286 | 3,431 | +50% | 0 | 0 | — |
case-09 | fail→pass | 17,410 | 11,522 | -34% | 1 | 1 | 0% | 2,361 | 3,411 | +44% | 0 | 0 | — |
case-10 | fail→fail | 13,197 | 2,505 | -81% | 1 | 1 | 0% | 1,629 | 1,584 | -3% | 0 | 0 | — |
case-11 | pass→pass | 18,720 | 15,862 | -15% | 1 | 1 | 0% | 2,589 | 3,324 | +28% | 0 | 0 | — |
case-12 | pass→pass | 7,196 | 11,249 | +56% | 1 | 1 | 0% | 1,496 | 2,284 | +53% | 0 | 0 | — |
case-13 | pass→pass | 13,254 | 10,108 | -24% | 1 | 1 | 0% | 1,616 | 2,114 | +31% | 0 | 0 | — |
case-14 | pass→pass | 7,768 | 3,486 | -55% | 1 | 1 | 0% | 1,533 | 1,843 | +20% | 0 | 0 | — |
case-15 | pass→pass | 13,938 | 15,216 | +9% | 1 | 1 | 0% | 2,611 | 4,352 | +67% | 0 | 0 | — |
case-16 | pass→pass | 9,831 | 17,521 | +78% | 1 | 1 | 0% | 1,712 | 3,373 | +97% | 0 | 0 | — |
case-17 | pass→pass | 12,517 | 12,071 | -4% | 1 | 1 | 0% | 1,101 | 2,111 | +92% | 0 | 0 | — |
case-18 | fail→pass | 21,311 | 23,378 | +10% | 1 | 1 | 0% | 2,768 | 4,215 | +52% | 0 | 0 | — |
case-19 | pass→pass | 13,816 | 4,820 | -65% | 1 | 1 | 0% | 1,389 | 1,866 | +34% | 0 | 0 | — |
case-20 | fail→pass | 7,918 | 5,246 | -34% | 1 | 1 | 0% | 1,304 | 1,937 | +49% | 0 | 0 | — |
case-21 | pass→pass | 16,428 | 9,153 | -44% | 1 | 1 | 0% | 1,808 | 2,617 | +45% | 0 | 0 | — |
case-22 | fail→pass | 13,515 | 3,124 | -77% | 1 | 1 | 0% | 1,500 | 1,650 | +10% | 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 19 counted toward the lift figure. The other 3 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 +18 percentage points is the difference between those two pass rates over the 19 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
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.