search-first

name	search-first
description	Research existing solutions before writing custom code
origin	pm-pilot
version	1.0.0
triggers	["before implementing anything new","search first","find existing solution"]

Search First

Mandatory research before writing any new code. Reuse beats reinvention.

When to Use

• Before implementing any new feature
• Before writing a utility function
• Before adding a new dependency
• Before building infrastructure (CI, deploy, tooling)

Procedure

Step 1: GitHub Code Search

# Search for existing implementations
gh search repos "{feature keywords}" --limit 5
gh search code "{function signature or pattern}" --limit 10

Look for:

• Battle-tested implementations with stars and activity
• Patterns that solve 80%+ of the problem
• Forkable/portable code with compatible licenses

Step 2: Package Registry Search

Check the relevant registry:

• npm: npm search {keywords}
• PyPI: pip index versions {package} or web search
• crates.io: cargo search {keywords}
• Go: pkg.go.dev search

Evaluate packages on:

• Download count and trend
• Last publish date (stale = risk)
• Dependency count (fewer = better)
• License compatibility

Step 3: Library Documentation

• Read official docs for candidate libraries
• Check API matches the use case
• Verify version compatibility with project

Step 4: Decision

Output one of:

REUSE: {package/repo} - {why it fits}
ADAPT: {package/repo} - {what needs modification}
BUILD: No suitable existing solution - {why}

Output Format

Search: {what was needed}

GitHub: 3 repos found, 1 strong match
  - github.com/user/repo (2.1k stars, MIT, active)

Registry: 2 packages evaluated
  - package-name (50k weekly downloads, v3.2.1)

Decision: REUSE package-name
Rationale: Covers all requirements, well-maintained, MIT license
Install: npm install package-name

Rules

• NEVER skip this step for non-trivial implementations
• Spending 5 minutes searching saves hours of coding
• A 90% solution from a library beats a 100% custom solution
• If BUILD is the decision, document why alternatives were rejected
• For trivial utilities (< 10 lines), skip registry search
• Always check license compatibility before adopting