GitHub Hunter Skill

Automatically discovers GitHub repositories relevant to BidDeed.AI and Life OS, scores them 0-100 based on criteria, and archives them to Supabase with integration recommendations.

Workflow

1. Discovery Phase

Search GitHub for repositories using relevant keywords extracted from:

• User requests ("find repos for X")
• Video transcripts (projects mentioned in content)
• Technology domains (foreclosure data, ADHD productivity, etc.)

2. Scoring Phase (0-100)

Score each repository on:

• Stars (0-25 points): Logarithmic scale, 10K+ stars = 25
• Recency (0-20 points): Last updated within 30 days = 20, 1+ year = 0
• Documentation (0-15 points): README quality, examples, API docs
• Relevance (0-25 points): Direct applicability to BidDeed.AI or Life OS
• License (0-15 points): MIT/Apache = 15, GPL = 10, Proprietary = 0

Formula:

score = min(100, 
    (log10(stars + 1) / log10(10000)) * 25 +
    max(0, 20 - (days_since_update / 15)) +
    documentation_score +
    relevance_score +
    license_score
)

3. Archive Phase

Insert to Supabase insights table with:

{
  "category": "github_discovery",
  "subcategory": "auto_hunter",
  "title": "GitHub Hunter: {repo_name}",
  "content": {
    "repo_url": "https://github.com/{owner}/{name}",
    "score": 85,
    "stars": 1234,
    "description": "...",
    "language": "Python",
    "license": "MIT",
    "last_updated": "2025-12-20",
    "integration_recommendation": "...",
    "relevant_to": ["biddeed", "life-os"]
  }
}

4. Alert Phase

Notify Ariel via response with:

• Repository name and URL
• Score (with color coding: 🟢 80+, 🟡 60-79, 🟠 40-59, 🔴 <40)
• Brief summary (1-2 sentences)
• Integration recommendation
• Direct action: "Add to {repo}?" with yes/no

Usage Triggers

Explicit requests:

• "Find GitHub repos for {topic}"
• "Search for projects about {domain}"
• "Discover repositories related to {technology}"

Context-aware triggers:

• Video transcripts mentioning GitHub projects → auto-hunt after transcript
• Articles/docs with GitHub URLs → extract and score
• User says "what could we integrate from that?" after discussing a topic

Scoring Examples

Score 95: fastapi/fastapi

• 75K stars (25), updated 2 days ago (20), excellent docs (15), highly relevant to BidDeed.AI API (25), Apache license (15)

Score 72: user/small-foreclosure-tool

• 45 stars (8), updated 1 week ago (18), basic README (8), perfect relevance (25), MIT (15)

Score 38: abandoned/old-project

• 500 stars (15), updated 2 years ago (0), no docs (0), tangential relevance (8), MIT (15)

Integration Recommendations Format

Provide actionable integration steps:

### Integration Recommendation

**To BidDeed.AI:**
1. Use {feature} for {existing_workflow_stage}
2. Replace {current_approach} with {repo_approach}
3. Add workflow: `.github/workflows/{new_workflow}.yml`

**To Life OS:**
1. Integrate {tool} for {productivity_feature}
2. Add skill: `.claude/skills/{skill_name}/`
3. Update orchestrator to call {function}

**Estimated effort:** {hours} hours
**Dependencies:** {list}
**Risk level:** {low/medium/high}

Alert Template

🔍 **GitHub Hunter Discovery**

**{repo_name}** [{score_emoji} {score}/100]
https://github.com/{owner}/{name}

{description}

**Stats:** ⭐ {stars} | 📅 {last_updated} | 📜 {license} | 💬 {language}

**Integration:**
{integration_recommendation}

**Add to BidDeed.AI?** [Yes/No]
**Add to Life OS?** [Yes/No]

Advanced: Batch Discovery

When user provides a list of topics or a domain:

# Example: "Find repos for foreclosure data scraping, PDF parsing, and workflow orchestration"

Topics:
1. foreclosure data scraping
2. PDF parsing  
3. workflow orchestration

For each topic:
- Search GitHub API with 3-5 keyword variations
- Score top 10 results per topic
- Archive scores 60+ to Supabase
- Alert Ariel with top 3 across all topics

Repository Addition Workflow

When user approves a repo:

1. Determine target repo:

   • BidDeed.AI → breverdbidder/biddeed-conversational-ai
   • Life OS → breverdbidder/life-os
   • Both → add to both

2. Create integration plan:

   • If library: Add to requirements.txt or package.json
   • If workflow: Create .github/workflows/{name}.yml
   • If skill: Create .claude/skills/{name}/SKILL.md
   • If script: Add to src/integrations/ or agents/

3. Document in README:

   • Add to "Integrations" section
   • Link to repo
   • Note version and license

4. Archive decision:

   • Update Supabase insight with integration_status: "added"
   • Record which repo(s) it was added to
   • Note commit SHA

Filters

Exclude repos with:

• Archived status
• No commits in 2+ years (unless legendary/foundational)
• Proprietary license for core BidDeed.AI features
• <10 stars AND <30 days old (likely spam)

Prioritize repos with:

• Python (BidDeed.AI), JavaScript/TypeScript (Life OS)
• AI/ML, web scraping, document processing, workflow automation
• Active maintenance (commits in last 60 days)
• Clear documentation
• Permissive licenses (MIT, Apache, BSD)

GitHub API Usage

Use web_search to find repos, then web_fetch for details:

# Search
query = "foreclosure auction data scraping language:python"
search_url = f"https://github.com/search?q={query}&type=repositories&s=stars&o=desc"

# Fetch repo details
repo_url = "https://api.github.com/repos/{owner}/{name}"
# Get: stars, last_updated, description, language, license, topics

Supabase Schema

Insert to insights table at mocerqjnksmhcjzxrewo.supabase.co:

INSERT INTO insights (category, subcategory, title, content, created_at)
VALUES (
  'github_discovery',
  'auto_hunter',
  'GitHub Hunter: {repo_name}',
  '{
    "repo_url": "...",
    "score": 85,
    "stars": 1234,
    "description": "...",
    "language": "Python",
    "license": "MIT",
    "last_updated": "2025-12-20",
    "integration_recommendation": "...",
    "relevant_to": ["biddeed"],
    "integration_status": "pending"
  }'::jsonb,
  NOW()
);

Example Session

User: "Find GitHub repos for PDF form filling and data extraction"

Claude: [triggers github-hunter skill]

1. Search: "PDF form filling python", "PDF data extraction", "fillable pdf automation"
2. Discover: 
   - PyPDF2 (8.2K stars, score: 78)
   - pdfplumber (6.1K stars, score: 82)
   - pdf-form-fill (234 stars, score: 71)
3. Archive top 2 to Supabase
4. Alert:

🔍 **GitHub Hunter Discovery**

**pdfplumber** [🟢 82/100]
https://github.com/jsvine/pdfplumber

Plumb a PDF for detailed information about tables, text, images. Maintained, excellent docs.

**Stats:** ⭐ 6.1K | 📅 2025-12-15 | 📜 MIT | 💬 Python

**Integration:**
Replace manual PDF parsing in BECA scraper with pdfplumber for structured table extraction.
Use for tax certificate downloads from RealTDM.

**Add to BidDeed.AI?** [Yes/No]

Notes

• Always archive to Supabase BEFORE asking for approval
• Score threshold for alerts: 60+
• Batch discovery: alert top 3 only, archive all 60+
• If repo already in our codebase, mark as integration_status: "existing"