Skills

A collection of skills for Claude Code, Codex, Cursor and OpenCode to supercharge your specification-driven development workflow!

Quick Start

npx skills add ikatsuba/skills

That's it! You're ready to go.

---

Installation Options

Install specific skills

Pick only what you need:

npx skills add ikatsuba/skills --skill spec:plan
npx skills add ikatsuba/skills --skill spec:research
npx skills add ikatsuba/skills --skill spec:design
npx skills add ikatsuba/skills --skill spec:breakdown
npx skills add ikatsuba/skills --skill spec:implement
npx skills add ikatsuba/skills --skill spec:test-plan
npx skills add ikatsuba/skills --skill spec:test
npx skills add ikatsuba/skills --skill git:commit
npx skills add ikatsuba/skills --skill git:amend
npx skills add ikatsuba/skills --skill git:changelog
npx skills add ikatsuba/skills --skill review:diff
npx skills add ikatsuba/skills --skill review:ux
npx skills add ikatsuba/skills --skill agent:design
npx skills add ikatsuba/skills --skill agent:eval
npx skills add ikatsuba/skills --skill agent:secure
npx skills add ikatsuba/skills --skill agent:review
npx skills add ikatsuba/skills --skill agent:prompt
npx skills add ikatsuba/skills --skill agent:tools
npx skills add ikatsuba/skills --skill agent:memory
npx skills add ikatsuba/skills --skill agent:workflow
npx skills add ikatsuba/skills --skill agent:rag
npx skills add ikatsuba/skills --skill agent:multi
npx skills add ikatsuba/skills --skill agent:context
npx skills add ikatsuba/skills --skill project:vision
npx skills add ikatsuba/skills --skill project:decompose
npx skills add ikatsuba/skills --skill project:status
npx skills add ikatsuba/skills --skill dev:create-skill

Global installation

Want skills everywhere? Go global:

npx skills add ikatsuba/skills -g

Browse available skills

npx skills add ikatsuba/skills --list

---

Available Skills

Project Orchestration

Skill	Description
project:vision	Define system goals, stakeholders, constraints, shared decisions
project:decompose	Break into specs with dependency graph and execution order
project:status	Dashboard showing progress across all specs

Specification Pipeline

Skill	Description
spec:plan	Plan requirements — user stories, acceptance criteria
spec:research	Technical research — solution alternatives with pros/cons
spec:design	Technical design — architecture diagrams, interfaces
spec:breakdown	Task breakdown — implementation plan with checkboxes
spec:implement	Implement tasks — execute all, next, or specific task
spec:review	Review — validate documents for quality and consistency
spec:test-plan	Test plan — generate manual test scenarios
spec:test	Execute tests — walk through all, next, or specific test

Git

Skill	Description
git:commit	Smart commit with Conventional Commits
git:amend	Amend last commit
git:changelog	Generate human-readable changelogs for product, marketing, and support teams

Code Review

Skill	Description
review:diff	Diff review with structured feedback before committing
review:ux	Detailed UX review — user flows, edge cases, and UI consistency

AI Agent Design

Skill	Description
agent:design	Comprehensive agent system design — orchestrates research across all areas
agent:eval	Evaluation system — failure modes, metrics, eval test suite
agent:secure	Security audit — lethal trifecta, sandboxing, access control, guardrails
agent:review	Full pattern review — scores all patterns with maturity assessment
agent:prompt	Prompt engineering — model selection, system prompt, few-shot examples
agent:tools	Tool design — schemas, MCP strategy, third-party integrations
agent:memory	Memory architecture — working memory, semantic recall, processors
agent:workflow	Workflow design — graph primitives, suspend/resume, streaming
agent:rag	RAG pipeline — chunking, embedding, vector DB, retrieval tuning
agent:multi	Multi-agent systems — org design, supervision, control flow
agent:context	Context engineering — parallelization, failure modes, compression

---

Workflow

These skills support a complete specification-driven development workflow:

Phase 0: Project Planning

For large systems with multiple specs:

1. /project:vision <name> — Creates .projects/<name>/vision.md
2. /project:decompose <name> — Creates .projects/<name>/specs.md
3. /project:status <name> — Dashboard across all specs (use anytime)

Phase 1: Specification

Build your blueprint:

1. /spec:plan <name> — Creates .specs/<name>/requirements.md
2. /spec:research <name> — Creates .specs/<name>/research.md
3. /spec:design <name> — Creates .specs/<name>/design.md
4. /spec:breakdown <name> — Creates .specs/<name>/tasks.md

Phase 2: Implementation

Let the magic happen:

5. /spec:implement <name> — Execute all tasks automatically
6. /spec:implement <name> next — Execute tasks one at a time with review
7. /spec:implement <name> <N> — Execute a specific task by number

Phase 3: Testing

Verify your implementation:

8. /spec:test-plan <name> — Generate manual test plan
9. /spec:test <name> — Walk through all pending tests
10. /spec:test <name> next — Execute one test at a time
11. /spec:test <name> <N> — Execute a specific test by number

Code Review

Review your code before committing:

/review:diff                # Auto-detect changes to review
/review:diff staged         # Review staged changes only
/review:diff branch         # Review all branch changes vs main
/review:diff src/auth/      # Review specific directory

UX Review

Review the user experience of a feature:

/review:ux user-auth         # UX review based on spec documents
/review:ux src/pages/settings/  # UX review of specific files
/review:ux branch            # UX review of all UI changes on branch

Changelog Generation

Keep your stakeholders in the loop:

/git:changelog last week           # Changes from the past 7 days
/git:changelog since 2024-01-01    # All changes since a date
/git:changelog v1.0.0..v1.1.0      # Changes between releases

Agent Design

Design AI agent systems from scratch or review existing ones:

/agent:design my-agent             # Full design with parallel research sub-agents
/agent:secure my-agent             # Security audit (lethal trifecta, guardrails)
/agent:eval my-agent               # Evaluation system (failure modes, test suite)
/agent:review my-agent             # Score against all patterns (0-66 points)

Or dive into a single area:

/agent:prompt                      # Prompt engineering and model selection
/agent:tools                       # Tool schemas, MCP, integrations
/agent:memory                      # Memory layers and processors
/agent:workflow                    # Workflow graph, streaming, observability
/agent:rag                         # RAG pipeline or alternatives
/agent:multi                       # Multi-agent organization and control flow
/agent:context                     # Context engineering and failure prevention

---

Documents Structure

.projects/
└── <project-name>/
    ├── vision.md           # System goals, constraints, shared decisions
    └── specs.md            # Spec list, dependency graph, execution order

.specs/
└── <spec-name>/
    ├── requirements.md     # User stories and acceptance criteria
    ├── research.md         # Research findings and chosen solutions
    ├── design.md           # Architecture, components, interfaces
    ├── tasks.md            # Implementation plan with checkboxes
    ├── test-plan.md        # Manual test plan with test cases
    ├── agent-design.md     # AI agent system design (from agent:design)
    ├── agent-eval.md       # Agent evaluation system (from agent:eval)
    └── agent-security.md   # Agent security audit (from agent:secure)

---

License

MIT