HarnessCommunity★ 2k

cc-sdd

Updated Jul 7, 2026

A spec-driven-development harness (v3) by gotalab supporting 8 AI coding agents. A /kiro-discovery entry point evaluates task scope and routes into: extend an existing spec, implement directly, create one spec, decompose into multiple specs, or mixed decomposition. /kiro-impl then runs each task with a fresh implementer (TDD RED→GREEN), an independent reviewer, and auto-debug on failure. Humans approve at phase gates. Inspired by the Kiro IDE spec methodology.

View on GitHub

What it does

/kiro-discovery
Evaluates a task or idea, writes brief.md (and roadmap.md for multi-spec work), and routes to the appropriate next command.
/kiro-impl
Autonomously implements tasks from tasks.md: fresh implementer per task (TDD RED→GREEN), independent reviewer, auto-debug on repeated failure.
/kiro-spec-requirements
Generates requirements.md with EARS-format requirements and acceptance criteria for a named spec.
/kiro-spec-design
Generates design.md with Mermaid architecture diagrams and a File Structure Plan that drives task boundaries.
/kiro-spec-tasks
Produces tasks.md with Boundary and Depends annotations per task, ready for /kiro-impl.
/kiro-spec-batch
Turns a roadmap into multiple specs in parallel, with cross-spec review to catch contradictions and interface mismatches.

Files (1)

OVERVIEW.mdprimary · markdown · 2.4 KB

## What it does

cc-sdd v3 is a spec-driven-development harness inspired by the Kiro IDE methodology. It installs 17 Agent Skills (loaded on demand) that cover a complete SDLC from discovery through autonomous implementation.

Entry point: /kiro-discovery evaluates a task or idea and routes into one of five paths — extend an existing spec, implement directly with no spec, create one new spec, decompose into multiple specs, or mixed decomposition. It writes brief.md (always) and roadmap.md (for multi-spec work), so you can resume a workstream without re-explaining scope.

Spec workflow: /kiro-spec-requirements → /kiro-spec-design (generates requirements.md and design.md with EARS-format acceptance criteria, Mermaid architecture diagrams, and a File Structure Plan) → /kiro-spec-tasks (produces tasks.md with Boundary and Depends annotations per task).

Autonomous implementation: /kiro-impl runs tasks from tasks.md one at a time. Each task gets a fresh implementer running TDD (RED→GREEN) behind a feature flag, an independent reviewer checking DoD compliance with file:line evidence, and an auto-debug pass that investigates root causes in a clean context when the reviewer rejects twice. Learnings from earlier tasks propagate forward via Implementation Notes in tasks.md.

Supports Claude Code (stable), Codex (stable), Cursor, Copilot, Windsurf, OpenCode, Gemini CLI, and Antigravity (all beta). 13 languages supported via --lang flag.

## Install

    npx cc-sdd@latest                      # Claude Code Skills (default)
    npx cc-sdd@latest --codex-skills       # Codex
    npx cc-sdd@latest --cursor-skills      # Cursor IDE
    npx cc-sdd@latest --lang ja            # Japanese docs

Then in your agent:

    /kiro-discovery Photo albums with upload, tagging, and sharing
    /kiro-spec-init photo-albums
    /kiro-spec-requirements photo-albums
    /kiro-spec-design photo-albums
    /kiro-spec-tasks photo-albums
    /kiro-impl photo-albums

## When to use

Use cc-sdd when spec boundaries are the hard problem — for example, a multi-developer feature where you need explicit interface contracts between modules before anyone writes code. The File Structure Plan and task Boundary annotations are what prevent one agent's implementation from stepping on another's. For smaller single-scope changes, /kiro-discovery will route directly to implementation, skipping the spec phase.