← Back to Discover

codebase-recon

Skill for coding agents that studies a repository's git history and reports where changes and bug fixes pile up, who holds the knowledge and how active development is, so you get oriented before opening any file.

Coding agents
Visit website ↗

What it does

codebase-recon is a skill that runs inside a coding agent. When you invoke /codebase-recon, it first checks how large the repository is, then analyses seven aspects of the git history in parallel: the files that change most, the files that attract bug fixes, how knowledge is spread among authors, team momentum, how often emergency fixes and reverts occur, recently added files, and active contributors. It then cross-references frequently changed files with bug-prone ones to point out the riskiest files, and finishes with a report and recommendations. The only requirements are git and a repository with commit history. Installation is through the skills.sh command, which the README says works with more than 20 coding agents, or through the Claude Code or Codex plugin systems. The project credits an article about git commands for codebase reconnaissance as its inspiration.

Capabilities

  • Scales the analysis to the size of the repository
  • Seven git-history dimensions analysed in parallel
  • Cross-references frequently changed files with bug-prone ones to flag risk
  • Installs as an agent skill or as a Claude Code or Codex plugin
  • Needs only git and commit history

When to use it

  • You are joining an unfamiliar codebase and want a map of its trouble spots before reading code.
  • You want to find files that change constantly and also attract fixes.
  • You want a quick view of knowledge concentration and development momentum.