GitHub Profile Rules Extraction

Operator Context

This skill operates as the main orchestration pipeline for extracting programming rules from a GitHub user's public profile. It implements a 7-phase pipeline that fetches data exclusively via the GitHub API (no git clone), analyzes code patterns across repos, extracts PR review comments for preference signals, compiles findings into deduplicated confidence-scored rules, and outputs actionable CLAUDE.md-compatible entries.

Hardcoded Behaviors (Always Apply)

• CLAUDE.md Compliance: Read and follow repository CLAUDE.md files before execution.
• API-Only Data Fetching: All GitHub data must be fetched via scripts/github-api-fetcher.py. No git clone, no subprocess git calls. This is a non-negotiable constraint.
• Rate Limit Awareness: Before each batch of API calls, check remaining quota. If --token is not provided, the unauthenticated limit is 60 req/hr.
• Evidence-Based Rules Only: Every generated rule must cite at least one repo or review where the pattern was observed. No generic advice.
• Confidence Scoring: Every rule gets a confidence level: high (3+ repos), medium (2 repos), low (1 repo).

Default Behaviors (ON unless disabled)

• Communication Style: Report progress by phase with counts (repos fetched, files sampled, patterns found, rules generated).
• Top-N Repos: Analyze the top 10 repos by stars/recent activity unless overridden by --max-repos.
• Review Priority: PR reviews given carry 2x weight vs code authored for preference signals.
• Output Dual Format: Always produce both CLAUDE.md markdown and JSON with confidence scores.