Skills

A collection of Claude Code skills for AI-driven development workflows.

Repository Layout

skills/          # one directory per skill, each with a SKILL.md
docs/            # guides about using the skills

Each skill lives at skills/<name>/SKILL.md. To install one, copy or symlink its directory into ~/.claude/skills/ (user-wide) or .claude/skills/ (project-only) — the skill directory itself, not the skills/ wrapper.

Available Skills

Specification & Planning

Skill	Description
write-spec	Writes and refines specs for AI-driven development. Guides discovery with targeted questions, then produces unambiguous requirements with testable verification criteria.
review-spec	Reviews a spec across ten quality dimensions (goal clarity, actors, inputs/outputs, failure modes, etc.) and produces a prioritized list of issues, warnings, and suggestions.
write-spec-plan	Generates an iterative implementation plan from a spec file and writes it to a companion -plan.md file. Requires the spec to have passed review-spec first.

Implementation & Verification

Skill	Description
implement-spec-plan	Executes a plan file one iteration at a time, verifying acceptance criteria and running full regression tests before advancing.
verify-spec	Confirms the codebase fully satisfies a spec by walking through each iteration's acceptance criteria and running the full test suite.
freeze-spec	Distills a completed spec into a compact, immutable "context capsule" — preserving purpose, public interfaces, key decisions, and known limits while dropping internal mechanics — for use as background context in future work.

Domain Modeling

Skill	Description
ddd-analysis	Produces a complete Domain-Driven Design document (ubiquitous language, bounded contexts, aggregates, domain events, etc.) from requirements or a product description, one section at a time with user confirmation.

Python Development

Skill	Description
python-coding	Standards and conventions for Python projects using uv, ruff, ty, strict typing, FastAPI, Strawberry, SQLAlchemy, Alembic, and Pydantic.
using-pytest	Guidelines for writing pytest tests following Given/When/Then/Clean patterns, with business-readable docstrings and organized test layers (unit, functional, interface, integration, evals).

AI Tooling

Skill	Description
write-system-prompt	Authors production-quality system prompts for LLM agents and assistants. Asks clarifying questions before drafting and structures output across role, instructions, constraints, output format, examples, and tools.

Typical Workflow

write-spec → review-spec → write-spec-plan → implement-spec-plan → verify-spec

New to spec-driven development? Read the Spec-Driven Development guide — an orientative walkthrough of the six spec skills, what each stage produces, and the mistakes beginners usually make.