snqb/github-quality-search

GitHub Quality Search Skill

Search GitHub for established, well-maintained libraries. Filters out low-quality repos automatically.

Use this skill when:

• Looking for libraries to replace custom implementations
• Need alternatives to current dependencies
• Want established solutions for common problems (auth, HTTP, caching, etc.)
• Evaluating library quality before adoption

Do NOT use this skill for:

• Finding specific projects you already know exist
• One-off code examples or gists
• Academic papers or research repos
• Your own private repositories

Quality Filters (Applied Automatically)

Minimum Bar:

• 100+ stars
• Commit in last 6 months
• Has documentation (README with usage examples)
• Has releases (not just random commits)

Red Flags (Auto-Excluded):

• Archived repositories
• One-person projects with <5 contributors
• No CI/CD setup
• No license
• Stale issues (>50 open issues with no recent activity)

Usage

from skills.github_quality_search import search_github

results = search_github(
    query="jwt authentication python",
    language="python",
    min_stars=100,
    topics=["authentication", "jwt"],
    max_results=5
)

for repo in results:
    print(f"{repo['name']} - {repo['stars']}⭐ - {repo['description']}")
    print(f"  Health: {repo['health_score']}/100")
    print(f"  Last commit: {repo['last_commit']}")
    print(f"  License: {repo['license']}")

Output Format

Each result includes:

• name: Full repo name (owner/repo)
• description: One-liner description
• stars: Star count
• language: Primary language
• topics: GitHub topics/tags
• last_commit: Days since last commit
• license: License type
• health_score: 0-100 (based on stars, activity, docs, CI)
• url: GitHub URL
• docs_url: Documentation URL (if exists)
• weekly_commits: Average commits per week (last 3 months)
• contributor_count: Total unique contributors

Health Score Calculation

health_score = (
    stars_score * 0.3 +        # Popularity (log scale)
    activity_score * 0.3 +     # Recent commits, issue response time
    docs_score * 0.2 +         # README quality, wiki, docs site
    community_score * 0.2      # Contributors, CI setup, license
)

Scores 80+ = Excellent, 60-79 = Good, 40-59 = Okay, <40 = Risky

Examples

Replace custom retry logic

results = search_github(
    query="http retry python",
    language="python",
    topics=["http-client", "retry"],
    min_stars=200
)
# Returns: requests, httpx, tenacity, backoff

Replace custom JWT handling

results = search_github(
    query="jwt token typescript",
    language="typescript",
    topics=["jwt", "authentication"],
    min_stars=500
)
# Returns: jsonwebtoken, jose, passport-jwt

Replace custom validation

results = search_github(
    query="data validation python",
    language="python",
    topics=["validation", "schema"],
    min_stars=1000
)
# Returns: pydantic, marshmallow, cerberus

API Rate Limits

GitHub API: 60 requests/hour (unauthenticated), 5000/hour (authenticated)

To use authenticated access:

pass insert github/personal-access-token
# Paste your GitHub PAT (Settings → Developer settings → Personal access tokens)

Skill automatically uses PAT from pass if available.

Dependencies

uv pip install requests python-dateutil

Anti-Patterns to Avoid

❌ Don't over-filter: If no results, relax constraints (reduce min_stars) ❌ Don't ignore health score: 500⭐ abandoned repo < 200⭐ active repo ❌ Don't skip license check: MIT/Apache-2.0 = safe, GPL = viral, no license = risky ❌ Don't cargo-cult: Just because it's popular doesn't mean it fits your use case

✅ Do compare alternatives: Run search multiple times with different queries ✅ Do check migration effort: Look at API surface area, breaking changes ✅ Do verify bundle size: For frontend libs, check bundlephobia.com ✅ Do read recent issues: 100+ open issues = maintenance burden