act

ai-coding-toolkit

A Claude Code plugin with AI coding skills, slash commands, and specialized agents.

Installation

Plugin (skills + commands + agents)

claude plugin install mguinada/ai-coding-toolkit

Skills only (via Agent Skills registry)

npx skills add mguinada/ai-coding-toolkit

Individual skills can also be installed from the monorepo:

npx skills add mguinada/ai-coding-toolkit --skill tdd

---

Skills

Skills are context-aware knowledge packages that activate automatically when their trigger conditions are detected in the conversation. No explicit invocation needed — Claude loads the relevant skill when the topic matches.

Skill	Activates when…
tdd	Writing tests, implementing features with TDD, Red-Green-Refactor
refactor	Simplifying code, reducing complexity, fixing code smells
git-commit	Creating commits, writing commit messages
create-pr	Opening pull requests, submitting changes for review
debug	Debugging errors, stack traces, failing tests, unexpected behaviour
docker	Dockerfiles, docker-compose, containerising applications
changelog	Updating CHANGELOG.md, preparing releases, writing release notes
release	Cutting releases, tagging versions, publishing packages
agents-md-creator	Creating or updating AGENTS.md / CLAUDE.md files
design-pattern-adopter	Applying GoF patterns, decoupling components, OO architecture
ai-engineering	Building agentic systems, choosing workflows vs. agents
prompt-engineering	Writing prompts for agents, reducing hallucinations
phoenix-observability	LLM tracing, evaluation, monitoring with Arize Phoenix
copilot-sdk	Embedding Copilot in apps, custom agents (Technical Preview)
rails	Ruby on Rails controllers, models, migrations, routing, jobs
typescript	TypeScript types, generics, tsconfig, monorepo setup
vitest	Vitest unit, component, and browser testing

See the full skills catalog for trigger phrases and scope details.

---

Slash Commands

Slash commands provide explicit, on-demand entry points into workflows. Type a command in Claude Code to trigger it directly.

Command	What it does
/act:ask [question]	Adversarial architectural consultation — four advisors (Systems Designer, Strategist, Scalability, Risk) analyse your question
/act:commit [hint]	Inspect staged changes, draft a conventional commit message, and execute the commit
/act:pr	Push the current branch and create a GitHub pull request with a conventional title, body, and labels
/act:changelog [version]	Generate a grouped CHANGELOG.md entry from git history since the last tag
/act:release [version]	Full release sequence: pre-flight checks → changelog → version bump → tag → GitHub release
/act:tdd [feature]	Drive the Red-Green-Refactor cycle with the tdd-guide agent enforcing write-tests-first
/act:refactor [file]	Apply safe, test-backed refactoring steps; uses tdd if new tests are needed
/act:debug [error]	Four-phase debugging: Reproduce → Locate → Fix → Verify; adds regression tests after the fix
/act:review [scope]	Parallel code review by code-reviewer, security-reviewer, python-reviewer (if .py files), database-reviewer (if DB changes)
/act:agents-md	Generate or update AGENTS.md / CLAUDE.md with progressive disclosure structure
/act:docker [service]	Scaffold a Dockerfile + .dockerignore, or audit and fix existing Docker configuration
/act:deploy-check [env]	Multi-layer deployment readiness check — tests, security scan, container audit, DB migration check → Go / No-Go verdict
/act:plan [feature]	Structured planning before implementation: clarify scope, iterate, yield control for review
/act:docs [scope]	Generate or update documentation (README, API reference, architecture guide, or tutorial)

Usage example:

/act:review src/auth/
/act:commit
/act:pr

---

Agents

Agents are specialized subagents available after plugin install. They are launched automatically by commands, or you can invoke them by name in a conversation.