skills/AGENTS.md

AGENTS.md

Project Overview

This repository maintains a curated set of Agent Skills for practical engineering work.

• Main content lives in skills/, with one folder per skill.
• Skills are Markdown-first artifacts built around a required SKILL.md file plus optional bundled assets.
• The repository is intentionally lightweight: formatting is the only automated check in CI.

Key technologies:

• Markdown for all skill instructions and references
• Bun for repository tooling
• Prettier for Markdown formatting checks
• GitHub Actions for CI formatting validation and optional formatting fixes

Repository Layout

Expected structure for each skill:

skills/
  <skill-name>/
    SKILL.md
    references/   # optional supporting docs
    scripts/      # optional executable helpers

Top-level files to keep in sync when relevant:

• README.md: English repository overview and skill catalog
• README.zh.md: Chinese repository overview
• .github/workflows/markdown.yml: Markdown formatting CI
• .github/workflows/markdown-fix.yml: manual workflow to format and commit Markdown changes

Setup Commands

Install the repository tooling:

bun install

Check Markdown formatting:

bun run check:md

Format Markdown files in place:

bun run format:md

Development Workflow

When adding or editing a skill:

1. Work inside the relevant folder under skills/<skill-name>/.
2. Keep the core instructions in SKILL.md.
3. Put large reference material in references/ instead of bloating SKILL.md.
4. Put reusable helper scripts in scripts/ when the workflow benefits from execution instead of long pasted commands.
5. Update the repository catalog in README.md when adding, renaming, or removing a skill.
6. Update README.zh.md when the change affects user-facing repository guidance there as well.

This repository follows progressive disclosure: agents should load bundled references and scripts only when a task actually calls for them.

Testing And Validation

There is no unit test suite in this repository today. The primary validation is Markdown formatting plus a careful review of skill structure and links.

Before finishing a change, run:

bun run check:md

If formatting fails, run:

bun run format:md

Then re-run:

bun run check:md

If you changed workflow or package metadata, make sure those files still align with the commands above:

• package.json
• .github/workflows/markdown.yml
• .github/workflows/markdown-fix.yml

Code Style And Authoring Conventions

• Follow .editorconfig: UTF-8, LF line endings, spaces, width 2.
• Markdown formatting is governed by .prettierrc.json, which preserves prose wrapping.
• Prefer clear, execution-oriented instructions over background exposition.
• Keep SKILL.md focused on when to use the skill and how to execute it.
• Avoid duplicating large documentation inside SKILL.md; move supporting material into references/.
• Keep examples and commands copy-pasteable.
• Use relative repository links for files in this repo.
• When linking to another repo file and the link text is the filename or path itself, format it as path/to/file.ext. If the link text is descriptive text rather than a filename, keep the descriptive label.
• Format bare URLs shown directly in prose as autolinks, for example <https://example.com>.

CI And Deployment

GitHub Actions runs Markdown formatting checks on pushes and pull requests that touch Markdown or formatting-related config.

• CI workflow: .github/workflows/markdown.yml
• Manual formatting workflow: .github/workflows/markdown-fix.yml

There is no application deployment pipeline in this repository.

Pull Request Guidelines

Before opening or updating a PR:

• Run bun run check:md.
• If you added a new skill, verify its folder matches the expected layout.
• Verify README catalog entries and links still resolve.
• Keep changes scoped to the skill or documentation you intended to modify.

Agent Notes

• Prefer small, targeted edits over broad rewrites across many skills.
• Do not introduce extra tooling unless the repository actually needs it; the current workflow is intentionally minimal.
• If you add scripts to a skill, keep them colocated under that skill's scripts/ directory.
• If a skill becomes large or complex, keep the instruction surface in SKILL.md concise and push details into bundled references.