Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/jorgeferrando/sdd-skills/llms.txt

Use this file to discover all available pages before exploring further.

SDD Skills is a collection of 18 slash-command skills for AI coding assistants that implement Spec-Driven Development (SDD) — a methodology where every code change starts with a specification, not code. Skills are compatible with Claude Code, Cursor, Codex (OpenAI), GitHub Copilot, and SkillKit-managed agents. Once installed, they guide the AI through a consistent, auditable pipeline from problem statement all the way to archived canonical specs, with every decision captured in an openspec/ directory that lives alongside your code.

​
The problem SDD solves

Without structure, AI-assisted development produces code that works today but drifts from intent over time. Three failure modes appear repeatedly:
  • Untracked changes — the AI makes implementation decisions that are never recorded, so the next session starts without context.
  • No audit trail — there is no link between a business requirement and the commit that fulfilled it.
  • Forgotten conventions — the AI applies patterns from its training data instead of your project’s established rules, causing review failures and technical debt.
SDD solves all three by making every decision explicit. Behavior is specified before it is designed. Design is approved before tasks are written. Tasks are committed atomically before verification. Nothing is implemented without being tracked.

​
High-level workflow

You describe what you want
        |
   AI follows the SDD workflow
        |
   spec -> design -> tasks -> code -> verify
        |
   Every decision is documented in openspec/
Skills are project-agnostic and LLM-agnostic. They work with any language, framework, or stack. Project-specific knowledge lives in openspec/steering/ files generated during setup — skills read those files to adapt their behavior to your project.

​
8-phase lifecycle

Every feature, bugfix, or refactor follows the same eight-phase cycle. Each phase produces a concrete artifact that captures the decisions made, so the next phase can start from a clean context.
PhaseSkillArtifactMode
1. Explore/sdd-explorenotes.mdInline
2. Propose/sdd-proposeproposal.mdInline (interactive)
3. Spec/sdd-specspecs/{domain}/spec.mdInline (interactive)
4. Design/sdd-designdesign.mdAgent (non-interactive)
5. Tasks/sdd-taskstasks.mdInline (interactive)
6. Apply/sdd-applycode + commitsAgent per task
7. Verify/sdd-verifyreport + PRAgent (non-interactive)
8. Archive/sdd-archivecanonical specsInline
All artifacts live under openspec/changes/{change-name}/. After archive, the delta specs are merged into openspec/specs/, keeping canonical specs permanently in sync with the codebase.

​
Compatibility note

SDD Skills works with any language, framework, and stack. The skills themselves contain only generic process instructions. Everything project-specific — your stack, your test commands, your naming rules, your architectural boundaries — goes into steering files that /sdd-init generates during setup. Skills read those steering files at runtime, so the same skill set adapts to a Django monolith, a Next.js app, a Go microservice, or anything else.

Install SDD Skills

Get SDD Skills running in Claude Code, Cursor, Codex, Copilot, or any SkillKit-managed agent.

Read the Quickstart

Walk through a complete SDD cycle — from project setup to your first archived change.

Build docs developers (and LLMs) love

Get started for freeTalk to us