Sandbox
@dceoy/speckit-agent-skills

Spec Kit skills for Claude Code, Codex, and Cursor

This repo tracks Spec Kit skills and the generated files that make them available in different agent runtimes. The shared `speckit-*` skills live in `skills/`, while tool-specific entry points are checked in under folders like `.claude/skills`, `.agents/skills`, `.cursor/skills`, and `.github/agents`.

132 stars28 forksShellUpdated 22d ago
Who it's for

Builders who want Spec Kit workflows available across multiple agent tools.

What it delivers

You can run the same Spec Kit process across different agents without rebuilding the setup by hand.

What it does

Shared Spec Kit skills

Includes the `speckit-*` skills for constitution, specify, baseline, clarify, plan, tasks, analyze, converge, implement, checklist, and tasks-to-issues.

Tool-specific runtime layouts

Checks in the generated files each runtime expects, including Claude Code, Codex CLI, Cursor, GitHub Copilot CLI, OpenCode, and Gemini CLI.

Spec Kit integration state

Stores manifests, scripts, templates, and init options under `.specify/` so generated outputs can be recreated consistently.

Version-checked regeneration

Tracks the current Spec Kit release in `VERSION.md` and regenerates outputs when the version changes.

Repository-specific utility skill

Includes `claude-command-converter` for turning Claude command content into Agent Skill format.

How to get it

  1. 1Clone this repository.
    git clone https://github.com/dceoy/speckit-agent-skills.git
    cd speckit-agent-skills
  2. 2Initialize a target project with the Spec Kit integration appropriate for your agent.
    specify init --here --integration <integration>
  3. 3If the target runtime consumes Agent Skills directly, copy the required skills from…
    cp -a skills/* /path/to/project/skills/

README

speckit-agent-skills

Agent skills and generated runtime entry points for Spec Kit.

Overview

This repository tracks the current Spec Kit output together with a small set of repository-specific skills.

  • Shared Agent Skills - skills/ contains the shared speckit-* skills plus repository-specific skills. Claude Code uses them through .claude/skills, while Codex CLI uses them through .agents/skills.
  • Cursor Agent - Spec Kit-generated skills live in .cursor/skills/.
  • GitHub Copilot CLI - Spec Kit-generated agents and prompt wrappers live in .github/agents/ and .github/prompts/.
  • OpenCode - Spec Kit-generated commands live in .opencode/commands/.
  • Gemini CLI - .gemini/commands/ remains checked in, but Gemini is not part of the current CI regeneration set.
  • Spec Kit infrastructure - .specify/ contains integration state, manifests, scripts, and templates used by the generated skills and commands.

The tracked Spec Kit version is recorded in VERSION.md. CI checks the latest Spec Kit release and regenerates outputs only when that version changes.

Current integration model

The repository CI currently runs specify init --integration for:

  • codex
  • agy
  • cursor-agent
  • copilot
  • opencode
  • claude

The generated files are therefore not all stored in the same layout:

Runtime / integrationRepository pathRole
Agent Skills (agy)skills/Shared Spec Kit skills
Claude Code.claude/skills -> ../skillsShared skills exposed to Claude Code
Codex CLI.agents/skills -> ../skillsShared skills exposed to Codex
Cursor Agent.cursor/skills/Cursor-specific generated skill copies
GitHub Copilot CLI.github/agents/, .github/prompts/Generated agents and prompt wrappers
OpenCode.opencode/commands/Generated commands
Gemini CLI.gemini/commands/Checked-in commands; not regenerated by the current CI configuration

Legacy layouts such as .claude/commands/, .codex/prompts/, and .opencode/command/ are intentionally not maintained.

Quickstart

  1. Clone this repository.

    git clone https://github.com/dceoy/speckit-agent-skills.git
    cd speckit-agent-skills
    
  2. Install Spec Kit.

  3. Initialize a target project with the Spec Kit integration appropriate for your agent.

    specify init --here --integration <integration>
    
  4. If the target runtime consumes Agent Skills directly, copy the required skills from skills/ into that project's skills directory when needed.

    cp -a skills/* /path/to/project/skills/
    
  5. Invoke the relevant speckit-* skill or runtime command from your agent.

Spec Kit workflow

The main Spec-Driven Development flow in this repository is:

  1. Constitution - Define project principles.
  2. Specify - Capture requirements for new work.
    • Or Baseline - Derive a specification from an existing codebase.
  3. Clarify (optional) - Resolve material ambiguities.
  4. Plan - Produce the implementation strategy.
  5. Tasks - Generate ordered work items.
  6. Analyze (optional) - Check consistency across the specification, plan, and tasks.
  7. Converge (optional) - Compare an existing implementation with the spec/plan/tasks and append remaining work to tasks.md.
  8. Implement - Execute the tasks.

Additional utilities include Checklist for requirements-quality checks and Tasks to Issues for converting generated tasks into GitHub issues.

Visual workflow

flowchart TD
  C0["speckit-constitution"] --> C1["speckit-specify"]
  C0 --> B["speckit-baseline"]
  C1 --> C2["speckit-plan"]
  B --> C2
  C2 --> C3["speckit-tasks"]
  C3 --> C4["speckit-implement"]

  C1 -.-> O1["speckit-clarify"]
  B -.-> O1
  O1 -.-> C2

  C3 -.-> O2["speckit-analyze"]
  O2 -.-> C4

  C3 -.-> O3["speckit-converge"]
  O3 -.-> C4

  C3 -.-> O4["speckit-taskstoissues"]

  C1 -.-> O5["speckit-checklist"]
  C2 -.-> O5
  C3 -.-> O5

Skills

Spec Kit skills

The current shared speckit-* set includes:

  • speckit-analyze
  • speckit-baseline
  • speckit-checklist
  • speckit-clarify
  • speckit-constitution
  • speckit-converge
  • speckit-implement
  • speckit-plan
  • speckit-specify
  • speckit-tasks
  • speckit-taskstoissues

Most Spec Kit-managed skills are tracked by the manifests in .specify/integrations/. speckit-baseline is a repository-specific skill for deriving specifications from existing code.

Utility skills

  • claude-command-converter - Repository-specific utility for converting Claude command content into Agent Skill format.

Structure

.
├── skills/                      # Shared and repository-specific Agent Skills
├── .agents/
│   └── skills -> ../skills      # Codex CLI access to shared skills
├── .claude/
│   └── skills -> ../skills      # Claude Code access to shared skills
├── .cursor/
│   └── skills/                  # Cursor Agent generated skills
├── .gemini/
│   └── commands/                # Checked-in Gemini commands
├── .github/
│   ├── agents/                  # GitHub Copilot generated agents
│   ├── prompts/                 # GitHub Copilot prompt wrappers
│   └── workflows/               # CI and Spec Kit regeneration workflows
├── .opencode/
│   └── commands/                # OpenCode generated commands
├── .specify/
│   ├── integrations/            # Integration manifests
│   ├── scripts/bash/            # Spec Kit helper scripts
│   ├── templates/               # Spec Kit templates
│   ├── init-options.json        # Last initialization options
│   └── integration.json         # Current integration state
├── .vscode/
│   └── settings.json            # Workspace settings
└── VERSION.md                   # Tracked Spec Kit version

Regeneration

.github/workflows/ci.yml calls .github/workflows/speckit-init.yml. The workflow:

  1. resolves the latest Spec Kit release,
  2. records specify --version in VERSION.md,
  3. skips regeneration when VERSION.md is unchanged,
  4. runs specify init --force --here --ignore-agent-tools --script sh --integration for the configured integrations,
  5. formats generated Markdown and JSON, and
  6. commits updated Spec Kit outputs when the tracked version changes.

Treat manifest-managed runtime files as generated output. Prefer changing the integration configuration or repository-specific skills instead of manually maintaining generated copies.

Prerequisites

Install and authenticate only the runtime tools you intend to use, plus Spec Kit itself.

  • Claude Code - Uses .claude/skills.
  • OpenAI Codex CLI - Uses .agents/skills.
  • Cursor Agent - Uses .cursor/skills/.
  • GitHub Copilot CLI - Uses .github/agents/ and .github/prompts/.
  • OpenCode - Uses .opencode/commands/.
  • Gemini CLI - Can use the checked-in .gemini/commands/, but those files are outside the current regeneration set.
  • Spec Kit - Install from github.com/github/spec-kit.

Usage notes

  • Skills do not always auto-run; use the runtime's skill invocation flow or request the skill explicitly.
  • If a skill fails, inspect its SKILL.md and verify the required Spec Kit project structure and prerequisites.
  • Run Spec Kit helper scripts from the repository root and prefer their structured output modes such as --json when available.
  • Do not restore legacy runtime layouts unless a currently supported Spec Kit integration starts generating them again.

License

See LICENSE for details.

Files in the repo

Repository payload13 top-level entries
  • .agents
  • .claude
  • .cursor
  • .gemini
  • .github
  • .opencode
  • .specify
  • .vscode
  • skills
  • .gitignore
  • LICENSE
  • README.md
  • VERSION.md

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More skills

obra/
superpowers

An agentic skills framework & software development methodology that works.

285k
1 add

Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.

117k
1 add
Vincentwei1021/
anything2explainer

Topic in, narrated explainer video out. A Claude Code / Codex skill that turns any topic into a black-canvas motion-graphics explainer video with TTS voiceover, subtitles and a chapter progress bar. Chinese or English; every frame drawn in code with Remotion.

666

Open-source AI job search: scan job portals, evaluate listings into a structured A-H report with a global 1-5 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)

71k