Unnamed Skill

Summarize an arbitrary code repo (git or non-git) into a small, high-signal facts packet: languages, build/test tooling clues, notable paths, and size stats.

$ 安裝

git clone https://github.com/WilletJiang/codex-context-kit-skills /tmp/codex-context-kit-skills && cp -r /tmp/codex-context-kit-skills/skills/repo_facts ~/.claude/skills/codex-context-kit-skills

// tip: Run this command in your terminal to install the skill


name: repo_facts description: Summarize an arbitrary code repo (git or non-git) into a small, high-signal facts packet: languages, build/test tooling clues, notable paths, and size stats.

repo_facts

Purpose

Produce a compact "facts packet" about the repository so the agent does not waste tokens repeatedly re-discovering:

  • What languages dominate
  • What build / test systems appear to be used
  • Where source code and generated artifacts likely live
  • Whether it's a git repo (and basic git state if available)

This skill is intentionally small: it avoids long README ingestion and avoids speculative instructions.

How to use

Run the script from the repo root (or pass a path):

repo_facts .
repo_facts . --no-tree
repo_facts /path/to/repo --json

# (fallback if wrappers are not on PATH)
CODEX_HOME="${CODEX_HOME:-$HOME/.codex}"
python3 "$CODEX_HOME/skills/repo_facts/scripts/repo_facts.py" /path/to/repo --tree

Output contract (high level)

  • Default: Markdown summary (human-scannable) + "Next look" file pointers.
  • --json: machine-readable JSON for programmatic consumption.

Notes

  • Works without git. If the directory is a git worktree, it prefers git ls-files for fast/accurate enumeration.
  • Skips common heavy dirs by default (e.g., .git/, node_modules/, target/, build/, .venv/).