Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use up-to-date library and framework docs via Context7 MCP instead of training data. Activates for setup questions, API references, code examples, or when the user names a framework (e.g. React, Next.js, Prisma).
.claude/skills/affaan-m-documentation-lookup/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | 84% | 0% |
| case-05 | ✓→✓ | = Same ✓ | 280% | 0% |
| case-06 | ✓→✓ | = Same ✓ | 71% | 0% |
| case-08 | ✓→✓ | = Same ✓ | 223% | 0% |
| case-09 | ✓→✓ | = Same ✓ | 133% | 0% |
当用户询问库、框架或 API 时,通过 Context7 MCP(工具 resolve-library-id 和 query-docs)获取最新文档,而非依赖训练数据。
/vercel/next.js)。当用户出现以下情况时激活:
当请求依赖于库、框架或 API 的准确、最新行为时,请使用此技能。适用于配置了 Context7 MCP 的所有环境(例如 Claude Code、Cursor、Codex)。
调用 resolve-library-id MCP 工具,参数包括:
Next.js、Prisma、Supabase)。在查询文档之前,必须获取 Context7 兼容的库 ID(格式为 /org/project 或 /org/project/version)。如果没有从此步骤获得有效的库 ID,请勿调用 query-docs。
从解析结果中,根据以下原则选择一个结果:
/org/project/v1.2.0)。调用 query-docs MCP 工具,参数包括:
/vercel/next.js)。限制:每个问题调用 query-docs(或 resolve-library-id)的次数不要超过 3 次。如果 3 次调用后答案仍不明确,请说明不确定性并使用您掌握的最佳信息,而不是猜测。
libraryName: "Next.js"、query: "How do I set up Next.js middleware?" 调用 resolve-library-id。/vercel/next.js)。libraryId: "/vercel/next.js"、query: "How do I set up Next.js middleware?" 调用 query-docs。middleware.ts 示例。libraryName: "Prisma"、query: "How do I query with relations?" 调用 resolve-library-id。/prisma/prisma)。libraryId 和查询调用 query-docs。include 或 select)并附上文档中的简短代码片段。libraryName: "Supabase"、query: "What are the auth methods?" 调用 resolve-library-id。| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 14,258 | 5,763 | -60% | 1 | 1 | 0% | 2,808 | 1,425 | -49% | 0 | 0 | — |
case-02 | fail→fail | 9,137 | 4,413 | -52% | 1 | 1 | 0% | 1,988 | 1,437 | -28% | 0 | 0 | — |
case-03 | fail→fail | 5,581 | 5,625 | +1% | 1 | 1 | 0% | 1,155 | 1,536 | +33% | 0 | 0 | — |
case-04 | fail→fail | 5,703 | 5,033 | -12% | 1 | 1 | 0% | 1,211 | 1,476 | +22% | 0 | 0 | — |
case-05 | pass→pass | 2,283 | 2,942 | +29% | 1 | 1 | 0% | 451 | 1,712 | +280% | 0 | 0 | — |
case-06 | pass→pass | 4,963 | 2,037 | -59% | 1 | 1 | 0% | 911 | 1,562 | +71% | 0 | 0 | — |
case-07 | fail→fail | 4,698 | 3,048 | -35% | 1 | 1 | 0% | 659 | 1,698 | +158% | 0 | 0 | — |
case-08 | pass→pass | 3,487 | 2,640 | -24% | 1 | 1 | 0% | 501 | 1,620 | +223% | 0 | 0 | — |
case-09 | pass→pass | 6,084 | 10,438 | +72% | 1 | 1 | 0% | 1,235 | 2,877 | +133% | 0 | 0 | — |
case-10 | pass→pass | 2,433 | 3,157 | +30% | 1 | 1 | 0% | 517 | 1,796 | +247% | 0 | 0 | — |
case-11 | fail→fail | 10,697 | 4,457 | -58% | 1 | 1 | 0% | 2,162 | 1,389 | -36% | 0 | 0 | — |
case-12 | fail→pass | 6,103 | 4,440 | -27% | 1 | 1 | 0% | 971 | 1,789 | +84% | 0 | 0 | — |
case-13 | fail→fail | 6,276 | 4,251 | -32% | 1 | 1 | 0% | 1,230 | 1,441 | +17% | 0 | 0 | — |
case-14 | fail→fail | 12,015 | 4,644 | -61% | 1 | 1 | 0% | 2,312 | 1,450 | -37% | 0 | 0 | — |
case-15 | fail→fail | 12,966 | 4,636 | -64% | 1 | 1 | 0% | 2,836 | 1,426 | -50% | 0 | 0 | — |
case-16 | fail→fail | 5,034 | 9,194 | +83% | 1 | 1 | 0% | 1,100 | 2,959 | +169% | 0 | 0 | — |
case-17 | fail→fail | 12,278 | 3,897 | -68% | 1 | 1 | 0% | 2,219 | 1,395 | -37% | 0 | 0 | — |
case-18 | pass→pass | 3,271 | 1,941 | -41% | 1 | 1 | 0% | 534 | 1,450 | +172% | 0 | 0 | — |
case-19 | fail→fail | 9,005 | 4,603 | -49% | 1 | 1 | 0% | 1,611 | 1,460 | -9% | 0 | 0 | — |
case-20 | pass→pass | 4,873 | 23,118 | +374% | 1 | 1 | 0% | 923 | 1,850 | +100% | 0 | 0 | — |
case-21 | pass→pass | 6,186 | 3,654 | -41% | 1 | 1 | 0% | 1,228 | 1,808 | +47% | 0 | 0 | — |
case-22 | pass→pass | 9,167 | 5,612 | -39% | 1 | 1 | 0% | 1,594 | 2,246 | +41% | 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 12 counted toward the lift figure. The other 10 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 +5 percentage points is the difference between those two pass rates over the 12 comparable cases. 1 case got worse with the skill loaded, and it is 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.