tech-docs

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

.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
├── 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:

# Search for specific topics
just search-docs "prefetchQuery" 5
just search-docs "HydrationBoundary" 3
just search-docs "authentication session" 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: just search-docs "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 Docs outdated?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

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

1. Run: just search-docs "HydrationBoundary" 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:

# Step 1: Search for relevant docs
just search-docs "<user question keywords>" 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
just fetch-docs <documentation-url> <tech-name>

Fetching New Documentation

# Fetch docs from any URL
just fetch-docs https://tanstack.com/query/latest/docs

# With custom name
just fetch-docs https://nextjs.org/docs nextjs

Priority

1. just 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)