AgentSkillsCN

tech-docs

在上网搜索之前先访问本地技术文档。在研究任何技术栈问题时优先使用。

SKILL.md
--- frontmatter
name: tech-docs
description: Access local technical documentation before searching the internet. Use FIRST when researching any tech stack question.
allowed-tools: Read, Glob, Grep, Bash

Technical Documentation Skill

ALWAYS search .docs/ FIRST before searching the internet!

Documentation Directory

code
.docs/
├── README.md           # Overview
├── nextjs.md           # Next.js 16 App Router
├── tanstack-query.md   # TanStack Query / React Query
├── better-auth.md      # Better Auth
├── gorm.md             # GORM ORM
├── gorilla-mux.md      # Gorilla Mux Router
├── goca.md             # Goca CLI
├── orval.md            # Orval API Client Generator
├── shadcn.md           # shadcn/ui Components
├── tailwind.md         # Tailwind CSS 4
├── kamal-deploy.md     # Kamal Deployment
└── logging.md          # Logging (zerolog + Pino)

Search Documentation (RECOMMENDED)

Use the semantic search tool to find relevant documentation:

bash
# Search for specific topics
make search-docs q="prefetchQuery" n=5
make search-docs q="HydrationBoundary" n=3
make search-docs q="authentication session" n=5

# Output is LLM-optimized with relevance scores

When to use search-docs:

  • •Looking for specific patterns or APIs
  • •Finding code examples for a concept
  • •Researching how to implement something
  • •The docs are large and you need targeted results

Workflow

  1. •

    Question about a technology? → First: make search-docs q="your question" to find relevant sections → Or: Read .docs/<tech>.md directly if you know the file

  2. •

    Docs not available? → Then search the internet → Save results in .docs/ for later

  3. •

    Docs outdated? → Update with latest info

Example Usage

code
User: "How does HydrationBoundary work in TanStack Query?"

1. Run: make search-docs q="HydrationBoundary" n=5
2. Review the top results with relevance scores
3. If needed, read full file: .docs/tanstack-query.md
4. If not found/incomplete:
   - WebFetch from TanStack Query docs
   - Save relevant info to .docs/tanstack-query.md
5. Answer the question

Tool Calling Pattern

When the user asks about tech stack topics, use this pattern:

bash
# Step 1: Search for relevant docs
make search-docs q="<user question keywords>" n=5

# Step 2: If results found, read the specific section
# The search output shows file and section info

# Step 3: If no results, fetch new docs
make fetch-docs url=<documentation-url> name=<tech-name>

Fetching New Documentation

bash
# Fetch docs from any URL
make fetch-docs url=https://tanstack.com/query/latest/docs

# With custom name
make fetch-docs url=https://nextjs.org/docs name=nextjs

Priority

  1. •make search-docs (semantic search, fast, targeted)
  2. •.docs/ files (local, fast, project-specific)
  3. •.claude/skills/ (project-specific patterns)
  4. •Internet (only if local not available)