AgentSkillsCN

repo-command-discovery

在陌生的仓库中发现测试、Lint、类型检查、构建，以及端到端测试的命令。当您对命令尚不熟悉或存在歧义时，或在运行测试之前，又或当 CI 失败、您亟需掌握正确的本地命令时，本指南将助您快速上手。本指南适用于 Python、Node/TS、React、Nest，以及混合型单体仓库等多种技术栈。

SKILL.md
--- frontmatter
name: repo-command-discovery
description: Discover test, lint, typecheck, build, and e2e commands in an unfamiliar repo. Use when commands are unknown or ambiguous, before running tests, or when CI fails and you need the right local commands. Works across Python, Node/TS, React, Nest, and mixed monorepos.

Repo Command Discovery

Overview

Find repo-specific commands with a portable POSIX script and confirm ambiguous cases with the user.

Workflow

1) Choose a repo root

  • •Prefer the workspace root (with .git, package.json, pyproject.toml, or Makefile).
  • •For monorepos, run from the workspace root first.

2) Run discovery script (always)

  • •On skill invocation, immediately run the wrapper script.
  • •If no path is provided, use the current working directory.
  • •If the skill is symlinked into a repo at .codex/skills, use:
    • •sh .codex/skills/repo-command-discovery/scripts/run.sh [path]
  • •Otherwise use the absolute path to this skill:
    • •sh <skills-root>/repo-command-discovery/scripts/run.sh [path]
    • •Common skills roots: $CODEX_HOME/skills, ~/.codex/skills

3) Interpret output

  • •Use the highest-confidence commands first.
  • •If confidence is low or multiple options exist, ask the user to confirm.
  • •If multiple package.json files exist, ask which package or workspace to target.

4) Cache results

  • •Record the chosen commands in your notes and use them consistently.

Script output

The script prints:

  • •ROOT, PACKAGE_MANAGER, CONFIDENCE
  • •FAST_TESTS, FULL_TESTS, LINT, TYPECHECK, BUILD, E2E
  • •NOTES about ambiguity or monorepo hints