cc-sdd: Long-running spec-driven implementation for AI coding agents

Turn approved specs into long-running autonomous implementation

One command installs an agentic SDLC workflow as Agent Skills: discovery, requirements, design, tasks, and autonomous implementation with per-task independent review. Works across 8 AI coding agents, with the same 17-skill set on each.

👻 Kiro-inspired. Similar spec-driven, agentic SDLC style as Kiro IDE. Existing Kiro specs remain compatible and portable.

What's new in v3.0

cc-sdd v3.0 is a rework around Agent Skills and long-running autonomous implementation.

• /kiro-discovery as the new entry point. Discovery routes new work into one of: extend an existing spec, implement directly with no spec, create one new spec, decompose into multiple specs, or mixed decomposition. It writes brief.md and, when needed, roadmap.md, so you can resume a workstream without re-explaining scope.
• /kiro-impl for long-running autonomous implementation. Each task gets a fresh implementer running TDD (RED → GREEN) behind a feature flag, an independent reviewer, and an auto-debug pass that investigates root causes in a clean context when the implementer is blocked or the reviewer rejects twice. Learnings from earlier tasks propagate forward via ## Implementation Notes in tasks.md. 1 task per iteration, safe to re-run after interruption.
• Boundary-first spec discipline. design.md now includes a File Structure Plan that drives task boundaries. Tasks carry _Boundary:_ and _Depends:_ annotations. Review and validation look for boundary violations, not just style issues.
• /kiro-spec-batch for multi-spec initiatives. Turn a roadmap into multiple specs in parallel, with cross-spec review to catch contradictions, duplicated responsibilities, and interface mismatches.
• Agent Skills across 8 coding agents. 17 skills per install, loaded on demand (progressive disclosure). Claude Code and Codex are stable; Cursor, Copilot, Windsurf, OpenCode, Gemini CLI, and Antigravity are in beta. No external dependencies; subagents are spawned through each platform's native primitive.

Full skills-mode workflow and /kiro-impl internals: Skill Reference.

Upgrading from v1.x or v2.x? See the Migration Guide.

Why cc-sdd?

cc-sdd treats the spec as a contract between parts of the system, not a master command document handed to the agent. Code remains the source of truth. Specs make the boundaries between parts of the code explicit so humans and agents can work in parallel without constant synchronization.

The bet: explicit contracts at the right granularity let AI-driven development at team scale move faster, not slower. Agents write the spec, humans approve the contract at phase gates, code is what ships.

Boundaries are not overhead. They are what lets you move freely inside while protecting the outside.

Full rationale, trade-offs, and when-to-use / when-not-to-use: Why cc-sdd? A philosophy note.

Quick Start

cd your-project
npx cc-sdd@latest

The default installs Claude Code Skills with English docs. To pick another agent or language:

npx cc-sdd@latest --codex-skills --lang ja      # Codex, Japanese
npx cc-sdd@latest --cursor-skills --lang zh-TW  # Cursor IDE, Traditional Chinese

Supports 8 AI coding agents (Claude Code and Codex stable; Cursor, Copilot, Windsurf, OpenCode, Gemini CLI, and Antigravity in beta) and 13 languages. See Supported Agents for the full list.

Then, in your agent:

/kiro-discovery <idea>

Not sure where to start? Start with kiro-discovery. It routes your request and tells you what command to run next.

Common workflows

Legacy /kiro:* command modes are still available (--claude, --cursor, etc.) but are deprecated. See the Migration Guide for the upgrade path.

For larger approved task sets, run kiro-impl to start autonomous implementation with per-task subagent spawn, independent review, and auto-debug on failure.

See It In Action

Example: build a new Photo Albums feature.

/kiro-discovery Photo albums with upload, tagging, and sharing
# discovery writes brief.md (and roadmap.md when multi-spec) and suggests the next command
/kiro-spec-init photo-albums
/kiro-spec-requirements photo-albums
/kiro-spec-design photo-albums
/kiro-spec-tasks photo-albums
/kiro-impl photo-albums
# autonomous: fresh implementer, independent reviewer, and auto-debug per task

Typical spec outputs (under 10 minutes):

• requirements.md: EARS-format requirements with acceptance criteria.
• design.md: architecture with Mermaid diagrams and a File Structure Plan.
• tasks.md: implementation tasks with boundaries and dependency annotations.

Then /kiro-impl runs the tasks autonomously with TDD (RED → GREEN) behind feature flags, an independent reviewer pass, and auto-debug on failure.

Supported Agents

All 8 skills variants ship the same 17-skill set. The difference is how much real-world usage each platform integration has seen.

"Beta" does not mean "missing features", the 17 skills and templates are identical across all 8 platforms. It means the platform integration (subagent spawn behavior, ergonomics, SKILL.md loading) has had less real-world usage than Claude Code and Codex, and edge cases may still surface. Please report issues if you hit any.

Advanced Installation

Skills mode (recommended)

npx cc-sdd@latest                     # Claude Code Skills (default)
npx cc-sdd@latest --claude-skills     # Claude Code Skills
npx cc-sdd@latest --codex-skills      # Codex Skills
npx cc-sdd@latest --cursor-skills     # Cursor IDE Skills (beta)
npx cc-sdd@latest --copilot-skills    # GitHub Copilot Skills (beta)
npx cc-sdd@latest --windsurf-skills   # Windsurf IDE Skills (beta)
npx cc-sdd@latest --opencode-skills   # OpenCode Skills (beta)
npx cc-sdd@latest --gemini-skills     # Gemini CLI Skills (beta)
npx cc-sdd@latest --antigravity       # Antigravity Skills (beta, experimental)

Legacy modes (deprecated)

npx cc-sdd@latest --claude         # Claude Code commands (use --claude-skills)
npx cc-sdd@latest --claude-agent   # Claude Code subagents (use --claude-skills)
npx cc-sdd@latest --cursor         # Cursor IDE commands (use --cursor-skills)
npx cc-sdd@latest --copilot        # GitHub Copilot prompts (use --copilot-skills)
npx cc-sdd@latest --windsurf       # Windsurf IDE workflows (use --windsurf-skills)
npx cc-sdd@latest --opencode       # OpenCode commands (use --opencode-skills)
npx cc-sdd@latest --opencode-agent # OpenCode subagents (use --opencode-skills)
npx cc-sdd@latest --gemini         # Gemini CLI commands (use --gemini-skills)
npx cc-sdd@latest --codex          # Codex (blocked, use --codex-skills)
npx cc-sdd@latest --qwen           # Qwen Code

Language

npx cc-sdd@latest --lang ja        # Japanese
npx cc-sdd@latest --lang zh-TW     # Traditional Chinese
npx cc-sdd@latest --lang es        # Spanish
# Supports: en, ja, zh-TW, zh, es, pt, de, fr, ru, it, ko, ar, el

Advanced options

# Preview changes before applying
npx cc-sdd@latest --dry-run

# Custom specs directory
npx cc-sdd@latest --kiro-dir docs

Customization

Edit templates and rules in {{KIRO_DIR}}/settings/ to match your team's workflow.

• templates/: document structure for requirements, design, tasks.
• rules/: AI generation principles and judgment criteria.

Common use cases: PRD-style requirements, API and database schemas, approval gates, JIRA integration, domain-specific standards.

Customization Guide has practical examples with copy-paste snippets.

Documentation

Related resources

Articles & presentations

• Kiroの仕様書駆動開発プロセスをClaude Codeで徹底的に再現した (Zenn, Japanese)
• Claude Codeは仕様駆動の夢を見ない (Speaker Deck, Japanese)

External resources

• Kiro IDE: enhanced spec management and team collaboration.
• Kiro's Spec Methodology: the original spec-driven development methodology.

License

MIT License