Skip to content

Latest commit

 

History

History

README.md

Agents

16 autonomous workflow agents that orchestrate skills and commands into end-to-end workflows. Each agent is a multi-phase process that loads relevant skills for domain knowledge and can invoke slash commands for structured sub-tasks.

Note: Agents are not published to skills.sh. Install them by cloning this repo and copying the agent directory into your project.


Agent Catalog

Agent Description Triggers
api Full API development — design, OpenAPI specs, endpoint implementation, tests, and documentation "build an API", "create API for", "design API", "generate OpenAPI spec", "add API endpoints"
backend Production backend services — architecture decisions, database design, service layers, and operational readiness "build a backend", "create a service", "design the architecture", "implement the backend", "create a microservice"
bootstrap Project scaffolding with opinionated structure, documentation, and tooling "bootstrap new project", "scaffold project", "create new project"
content Professional written content — docs, blog posts, social media, emails, guides, and technical writing "write a blog post", "create documentation", "draft an email", "write content for", "create a README"
debugging Systematic error diagnosis — stack trace parsing, hypothesis formation, incremental fix verification "debug this", "why is this failing", "fix this error", "investigate this bug", "diagnose this issue"
deployment CI/CD pipelines, container builds, cloud deployment, environment promotion, and monitoring setup "deploy", "set up CI/CD", "create pipeline", "containerize", "set up GitHub Actions"
design-system Design token extraction from codebases or reference sites, Tailwind + CSS variable generation "extract design system", "analyze design tokens", "create design system from"
development Docs-first feature development — planning, spec creation, ADRs, and implementation tracking "new feature", "develop feature", "add feature", "implement feature"
extraction Pattern mining from codebases into reusable skills and documentation "extract patterns from", "analyze this project", "extract from this repo", "capture patterns"
frontend Production frontend interfaces — component design, styling, accessibility, and performance "build a component", "create a page", "design the UI for", "build the frontend", "create a landing page"
marketing Marketing pages, copy, conversion funnels, CRO analysis, and measurement setup "create marketing page", "write copy for", "optimize conversions", "marketing strategy", "CRO audit"
migration Multi-step migrations with rollback planning — database schemas, framework upgrades, architecture changes "migrate", "upgrade", "migration plan", "migrate database", "zero-downtime migration"
performance Performance optimization — bundles, runtime, memory, queries, API response times, Core Web Vitals "optimize performance", "why is this slow", "reduce bundle size", "fix memory leak", "profile this"
refactoring Systematic code refactoring — smell detection, planning, incremental execution with test verification "refactor this", "clean up this code", "improve code quality", "restructure", "reduce complexity"
refinement Consolidate extracted skills from staging into production-ready, project-agnostic patterns "refine staged content", "consolidate skills", "process staging"
testing Comprehensive test suites — strategy, generation, coverage analysis, and testing infrastructure "write tests for", "test this", "improve test coverage", "add tests", "testing strategy"

Installation

Step 1: Clone this repo

git clone https://github.com/wpank/ai ~/.ai-skills

Step 2: Copy agents into your project

Cursor (per-project)

From your project root:

# Single agent
mkdir -p .cursor/agents
cp -r ~/.ai-skills/agents/api .cursor/agents/api

# All agents
mkdir -p .cursor/agents
cp -r ~/.ai-skills/agents/* .cursor/agents/

Cursor (global)

mkdir -p ~/.cursor/agents
cp -r ~/.ai-skills/agents/api ~/.cursor/agents/api

Claude Code (per-project)

From your project root:

mkdir -p .claude/agents
cp -r ~/.ai-skills/agents/api .claude/agents/api

Claude Code (global)

mkdir -p ~/.claude/agents
cp -r ~/.ai-skills/agents/api ~/.claude/agents/api

Each agent directory contains AGENT.md (the workflow definition) and README.md. For best results, also install the skills each agent references — see the agent's README for the list.


How Agents Work

Each agent is defined in agents/<name>/AGENT.md and follows a consistent pattern:

  1. Skills — The agent loads relevant skills from skills/ for domain knowledge (listed in "Before Starting").
  2. Commands — The agent can invoke slash commands from commands/ for structured sub-workflows.
  3. Workflow — A multi-phase process with checkpoints and quality gates (design, implementation, verification, etc.).

When you install an agent as a rule or skill, you're giving the AI the full AGENT.md so it can follow that workflow. For best results, also install the skills that agent references (see each agent's README for the list).


Model Tiers

Each agent declares per-phase model tier recommendations in its YAML frontmatter (models:). This allows different phases to use different model tiers for cost/speed.

Phase type Typical tier Rationale
Discovery / assessment fast File reading, scanning, structure analysis
Planning / design reasoning Architecture decisions, tradeoff analysis
Execution / implementation standard Code generation with patterns
Verification / completion fast Running checks, formatting output

See Also

Resource Path Description
Skills catalog ../skills/ 115 skills by category
Commands reference ../commands/ 48 slash commands