Agent Skills

Personal collection of agent skills for Claude Code, claude.ai, and the Claude API.

Skills follow the Agent Skills format.

Available Skills

aws-cli

Run AWS CLI commands safely. Enforces identity verification, scoped queries with JMESPath, region/profile hygiene, and confirmation gates on destructive ops. Includes per-service references for S3, EC2, IAM, Lambda, and CloudWatch Logs.

Use when:

• Invoking aws ... for any service
• "List S3 buckets", "describe EC2 instances", "tail Lambda logs", "what AWS account am I in"
• Reviewing or proposing AWS infrastructure changes from the CLI

architecture-satisfaction

Run a bounded architectural refactor loop. Defines what satisfactory means, makes one reviewable architecture change at a time, live-tests affected behavior, runs an independent review, commits verified checkpoints, and records progress in /tmp/refactor-{projectname}.md.

Use when:

• Refactoring toward a stated architecture or untangling module boundaries
• Fixing dependency direction, code organization, or long-running architectural drift
• "Refactor until the architecture is right", "run the architecture satisfaction loop", "checkpoint this refactor as we go"

weekly-composer-dependency-audit

Run a weekly Composer dependency audit for PHP projects. Checks outdated direct dependencies, runs composer audit, groups updates by risk, and produces a safe upgrade summary.

Use when:

• Auditing Composer or PHP dependencies
• "Run composer audit", "check outdated Composer packages", "review Laravel package upgrades"
• Planning safe PHP dependency upgrades without applying changes yet

grill-me-plus

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Uses Claude Code's AskUserQuestion or Codex's request_user_input when available, with tool-specific schemas, recommended options, trade-offs, and fallbacks.

Use when:

• Stress-testing a plan or design before building
• "Grill me on this", "poke holes in my approach", "interview me about this plan"
• Resolving interdependent design decisions one branch at a time in Claude Code, Codex, or chat fallback

housekeeper

Run a conservative code-project housekeeping pass. Finds one proven low-risk cleanup at a time, makes the smallest coherent change, verifies behavior, and defers uncertain or approval-required candidates.

Use when:

• Cleaning dead code, stale files or comments, unused dependencies, duplicated logic, broken links, inconsistent names, or confusing structure
• "Run housekeeper", "run the housekeeper loop", "do a safe repository cleanup", "find low-risk cleanup opportunities"
• Improving project hygiene while protecting unrelated, uncommitted, generated, active, or uncertain work

repository-cleanup

Audit and clean Git repository state. Inventories branches, pull requests, commits, stashes, and worktrees; recovers valuable work; removes only proven stale state; and repeats until remaining repo state is intentional.

Use when:

• Cleaning stale local or remote branches, old pull requests, unmerged commits, forgotten stashes, or obsolete worktrees
• "Run repository cleanup", "clean up old branches", "audit worktrees", "recover abandoned branch work"
• Organizing Git state while protecting uncommitted changes, unpushed commits, dirty worktrees, and uncertain ownership

laravel-herd-worktrees

Run multiple branches of a Laravel project in parallel via git worktrees, each served by Laravel Herd at its own *.test URL. Helper scripts handle worktree creation, .env rewrites (APP_URL, DB, cache/session/redis prefixes), per-worktree DB provisioning, dependency install, and clean teardown. Includes Herd CLI reference, isolation rationale, and DB strategy guide.

Use when:

• Creating a worktree for a Laravel app served by Herd
• Debugging cross-branch .env/cache/session/DB collisions
• Wanting per-branch URLs like myapp-feat.test without manual config
• Running the same Laravel project on multiple PHP versions simultaneously

weekly-npm-dependency-audit

Run a weekly npm dependency audit for JavaScript and TypeScript projects. Checks outdated npm packages, runs npm audit when lockfile data is available, groups updates by risk, and produces a safe upgrade summary.

Use when:

• Auditing npm, JavaScript, or TypeScript dependencies
• "Run npm audit", "check outdated npm packages", "review package.json upgrades"
• Planning safe npm dependency upgrades without applying changes yet

nightly-docs-sweep

Run a documentation sweep for a codebase. Compares docs against the current implementation, updates stale material, verifies commands and examples, and leaves a reviewable change set or pull request.

Use when:

• Auditing documentation drift after implementation changes
• Refreshing READMEs, setup guides, API references, examples, architecture notes, or runbooks
• "Run an overnight docs sweep", "make docs match the code", "open a docs cleanup PR"

Installation

npx skills (any agent)

Install the whole collection:

npx skills add leek/agent-skills

Install a single skill:

npx skills add https://github.com/leek/agent-skills --skill aws-cli

Re-run either command to update.

Claude Code (plugin marketplace)

/plugin marketplace add leek/agent-skills
/plugin install leek-skills@leek-agent-skills

Claude Code (manual copy)

cp -r skills/<skill-name> ~/.claude/skills/

claude.ai

Upload the skill folder via skill settings, or paste the SKILL.md contents into a project's knowledge.

Claude API

See the Skills API Quickstart.

Skill Structure

Each skill is a folder under skills/:

skills/
  <skill-name>/
    SKILL.md         # required — frontmatter + instructions
    scripts/         # optional — helper scripts
    references/      # optional — supporting docs

The SKILL.md frontmatter requires only two fields:

---
name: my-skill
description: One sentence — what it does and when Claude should use it.
---

# My Skill

Instructions here.

See template/SKILL.md.example for a starting point and AGENTS.md for full conventions.

Creating a Skill

1. Copy template/SKILL.md.example to skills/<skill-name>/SKILL.md.
2. Fill in SKILL.md — keep it under 500 lines; offload reference material to sibling files.
3. Add the skill to .claude-plugin/marketplace.json under the leek-skills plugin.
4. Update the Available Skills section above.
5. Open a PR.

License

MIT