setup-ai-project-tooling

Files (1)

• SKILL.md

SKILL.md (first lines)

# Setup AI Project Tooling

Use this skill when a repository needs the standard files and folders that make AI-agent collaboration predictable: `AGENTS.md`, `PLAN.md`, `CONTEXT.md`, and project-local `skills/`.

## Inspect First

Before creating files, inspect the repository:

- Check `git status --short`.
- Read existing agent files such as `AGENTS.md`, `CLAUDE.md`, `.cursorrules`, `.windsurfrules`, `README.md`, `CONTEXT.md`, and `PLAN.md`.
- Identify the main stack, package manager, test commands, build commands, formatter, linter, CI provider, deployment target, and repository conventions.
- Preserve existing user instructions. Do not overwrite agent files without merging their content.

## Create Or Update Files

Create or update these files at the project root:

- `AGENTS.md`: operational rules for AI agents working in this repository.
- `CONTEXT.md`: durable project context, vocabulary, architecture notes, commands, and important constraints.
- `PLAN.md`: current implementation plan, milestones, acceptance criteria, and open questions.

Create this folder when project-local skills are useful:

```text
skills/
```

Each project-local skill should live in:

```text
skills/<skill-name>/SKILL.md
```

Use a project-local skill only for reusable behavior that is specific enough to help future agents, such as release workflow, framework conventions, migration rules, testing strategy, or domain-specific review criteria.

## AGENTS.md Contents

Include:

- Repository purpose and primary workflows.
- How to run tests, linting, formatting, type checks, builds, and local servers.
- Coding conventions and architecture boundaries.
- Git rules, including files or branches agents must not rewrite.
- Dependency and package manager rules.
- Documentation expectations.
- Verification expectations before final answers.

Keep instructions direct and enforceable. Avoid generic AI etiquette.

## CONTEXT.md Contents

Include:

- Product or project summary.
- Key domain terms.
- Architecture overview.
- Important directories and ownership boundaries.
- External systems, APIs, data sources, and credentials expectations.
- Known risks, invariants, and non-obvious constraints.
- Historical decisions that affect current implementation.

Do not put secrets in `CONTEXT.md`.

## PLAN.md Contents

Include:

- Goal.
- Scope and non-goals.
- Milestones or ordered tasks.
- Acceptance criteria.
- Validation commands.
- Open questions and blockers.

Keep `PLAN.md` useful for the current work. Move completed durable knowledge into `CONTEXT.md`.

## Project Skills

For each local skill:

- Add `SKILL.md` with frontmatter containing `name` and `description`.
- Make `name` match the directory name.
- Write the description as activation guidance.
- Keep instructions specific to this repository.

Example:

```yaml
---
name: repo-release-workflow
description: Use when preparing a release for this repository, including versioning, checks, and deployment notes.
---
```

## Validation

After setup:

- Run the repository's normal validation commands if available.
- Confirm the new files are tracked or intentionally ignored.
- Re-read `AGENTS.md`, `CONTEXT.md`, and `PLAN.md` for contradictions.
- Report what was created, what was merged, and any assumptions that still need owner confirmation.

From Guillaume-Lombardo/skills, MIT.

More from Guillaume-Lombardo/skills