AgentRepoCoach — AI agent codebase health scoring

Version updated for https://github.com/WouterDeBot/AgentRepoCoach to version v0.3.0.

  • This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

AgentRepoCoach is a GitHub Action and CLI tool that evaluates a codebase’s readiness for autonomous AI agents by computing a “Codebase Agent Health” (CAH) score, ranging from 0 to 100. It automates the analysis of five key areas—navigability, error quality, decision queryability, test quality, and module hygiene—identifying areas for improvement and enabling teams to maintain high standards for AI agent compatibility. The action can also enforce quality thresholds in CI pipelines, generating detailed reports in various formats for further insights.

What’s Changed

[0.3.0] — 2026-04-23

Added

  • compare CLI subcommand for local score file comparison (base vs. PR JSON reports)
  • PR bot module (pr_bot.py) for structured score comparison on pull requests
  • --compare flag on the default score command for inline baseline comparison
  • GitHub Actions workflow for automated PR score comments (cah-score.yml)
  • GitHub Actions CI pipeline with Python 3.11/3.12/3.13 test matrix (ci.yml)
  • CLI integration tests (test_cli_compare.py)

Fixed

  • Language detection priority when multiple adapters tie on confidence — tiebreaker now uses production file count to select the dominant language deterministically

Security

  • ReDoS regex safety guard (regex_safety.py) for user-configurable patterns — detects nested quantifiers before compilation to prevent catastrophic backtracking

Install: pip install agentrepocoach==0.3.0

Full changelog: https://github.com/WouterDeBot/AgentRepoCoach/compare/v0.2.0...v0.3.0