Research Skill

Comprehensive research capabilities combining codebase analysis and external documentation scraping. Provides targeted context for implementation tasks.

When to Use

Invoke this skill when:

• Starting implementation of a new feature
• Need to understand existing code patterns
• Require external library documentation
• Context priming for /tasks-autopilot

Part 1: Codebase Analysis

Process

Step 1: Analyze Task Requirements

From the task description, identify:

• What needs to be created (new files)
• What needs to be modified (existing files)
• What patterns to follow (existing code)

Task: "Implement password hashing for user registration"

Needs:
  Create: password hashing utilities
  Modify: user schema (add passwordHash)
  Follow: existing auth patterns

Step 2: Search for Relevant Files

By filename:

find src -name "*password*" -o -name "*auth*" -o -name "*user*" 2>/dev/null

By content:

grep -rl "password\|hash\|bcrypt" src/ --include="*.ts" --include="*.tsx"

By imports:

grep -rh "from.*auth\|from.*user" src/ --include="*.ts" | head -10

Step 3: Analyze Dependencies

For each found file:

What does it import?

grep "^import" src/auth/index.ts

What imports it?

grep -rl "from.*auth" src/ --include="*.ts"

Step 4: Detect Project Patterns

Testing framework:

grep -E "vitest|jest|mocha" package.json
ls src/__tests__/ 2>/dev/null || ls src/**/*.test.ts 2>/dev/null

API framework:

grep -rh "from.*hono\|from.*express\|from.*trpc" src/ --include="*.ts" | head -3

Database ORM:

grep -rh "from.*drizzle\|from.*prisma" src/ --include="*.ts" | head -3

Step 5: Score and Rank Files

Keep top 20 files within token budget.

Step 6: Generate Codemaps

For files that need modification, use /codemap:

/codemap src/db/schema/users.ts
/codemap src/server/routers/auth.ts

Pattern Detection Reference

Part 2: Documentation Scraping

Process

Step 1: Identify Libraries

From the task description, detect which libraries are involved:

Task: "Implement password hashing for user registration"

Libraries detected:
  - bcrypt (password hashing)
  - drizzle (database storage)

Step 2: Load Doc Sources Registry

cat .claude/lib/doc-sources.json

Step 3: Match Topics

For each library, find relevant topics:

bcrypt:
  - usage (matches: hash, password)

drizzle:
  - insert (matches: create user)
  - schema (matches: passwordHash field)

Step 4: Scrape with Exact Paths

Use mycli scrapers with exact URL and CSS selector:

# Scrape bcrypt docs
mycli scrape crawl4ai \
  --url "https://www.npmjs.com/package/bcrypt" \
  --selector "#readme"

# Scrape drizzle insert docs
mycli scrape crawl4ai \
  --url "https://orm.drizzle.team/docs/insert" \
  --selector "article"

Scraper Selection

Token Budget

• Max 500 tokens per topic
• Extract code examples preferentially
• Skip navigation, headers, footers

Output Format

{
  "task": "Implement password hashing",
  "codebase": {
    "files": {
      "toCreate": [
        { "path": "src/auth/password.ts", "purpose": "Password utils" }
      ],
      "toModify": [
        { "path": "src/db/schema/users.ts", "codemap": "..." }
      ],
      "toReference": [
        { "path": "src/server/routers/auth.ts", "codemap": "..." }
      ]
    },
    "patterns": {
      "testing": { "framework": "vitest", "location": "src/__tests__/" },
      "api": { "framework": "tRPC", "location": "src/server/routers/" },
      "database": { "orm": "drizzle", "location": "src/db/schema/" }
    }
  },
  "docs": [
    {
      "library": "bcrypt",
      "topic": "usage",
      "url": "https://www.npmjs.com/package/bcrypt",
      "content": "..."
    }
  ]
}

Integration with Context Priming

This skill is invoked by /tasks-prime:

/tasks-prime ENG-102
    ↓
1. Invoke research skill
    ↓
2. Codebase analysis
   - Find relevant files
   - Detect patterns
   - Generate codemaps
    ↓
3. Documentation scraping
   - Detect libraries
   - Scrape relevant topics
    ↓
4. Output context JSON

Adding New Libraries

Edit .claude/lib/doc-sources.json:

{
  "new-library": {
    "name": "New Library",
    "baseUrl": "https://docs.newlibrary.dev",
    "keywords": ["newlib", "specific-feature"],
    "topics": {
      "getting-started": {
        "path": "/docs/start",
        "selector": "article.content",
        "keywords": ["start", "install", "setup"]
      }
    }
  }
}