{
  "markdown": "# DDW — Decision-Driven Workflow\n\n**Ship code that never drifts.\nNo decision, no code. No task, no implementation.**\n\n> A system that prevents bad development, not just fixes it.\n\nDDW is not a PM tool.\nIt's a **development harness that enforces quality at the system level.**\n\n---\n\n## What it does\n\n- **Stops bad code before it exists**\n  Decisions → Tasks → Code. No shortcuts.\n\n- **Makes every change provable**\n  Small tasks with strict acceptance criteria.\n\n- **Enforces rules, not suggestions**\n  Hard hooks block violations. You can't \"prompt\" your way around it.\n\n- **Keeps spec and code in sync**\n  Drift detection catches inconsistencies automatically.\n\n- **Turns development into a system**\n  Decisions, tasks, QA, and retros — all logged, all connected.\n\n---\n\n## The Flow\n\nDecision → Task → Implement → QA → Integrate → Close\n\nOr full power:\n\n```\n/ddw:ideate → /ddw:decision → /ddw:task → /ddw:sendit → /ddw:review\n                                                            ↓\n                                                       /ddw:close\n                                                            ↓\n                                                     /ddw:prd close\n```\n\n---\n\n## Why it matters\n\nMost teams don't fail because they can't code.\nThey fail because:\n\n- Decisions are vague\n- Tasks are unclear\n- QA is inconsistent\n- Specs drift from reality\n\nDDW fixes all of that — by **making it impossible to do it wrong.**\n\n---\n\n## Enforcement\n\nTwo layers:\n\n- **Soft** — AI guidance (`CLAUDE.md`)\n- **Hard** — Shell hooks that physically block invalid actions\n\nIf it breaks the rules, **it doesn't run.**\n\n---\n\n## Positioning\n\n- PM tools manage work\n- CI tools test code\n- **DDW enforces how development happens**\n\n---\n\n## Install\n\n```json\n// .claude/settings.json\n{\n  \"plugins\": [\n    { \"path\": \"/path/to/ddw\" }\n  ]\n}\n```\n\nThen run `/ddw:init`.\n\n---\n\n## Commands\n\nEverything is a slash command. The bash scripts under `scripts/` are implementation details — skills wrap them.\n\n| Command | Purpose |\n|---|---|\n| `/ddw:init` | Bootstrap DDW into a project |\n| `/ddw:ideate` | Shape rough ideas into a PRD |\n| `/ddw:decision` | Create decision with architect review |\n| `/ddw:prd close PRD-id` | Close a PRD once its decisions exist |\n| `/ddw:task` | Break decision into scoped tasks |\n| `/ddw:sendit` | Auto-creates a per-task worktree, implements, runs review |\n| `/ddw:qa` | Automated QA: acceptance criteria + invariants |\n| `/ddw:review` | QA + tests + owner checklist |\n| `/ddw:pr` | (team-PR mode) Push branch, open GitHub PR, transition task to in_review |\n| `/ddw:audit` | Adversarial security audit (OWASP+STRIDE) on Opus. Standalone. |\n| `/ddw:close` | Spec update, drift, retro, archive, **remove worktree** |\n| `/ddw:sync-spec` | Update CURRENT_SPEC.md after a spec-affecting task lands (post-merge). Auto-invoked from `/ddw:close` when relevant. |\n| `/ddw:drift` | Check spec-code consistency |\n| `/ddw:architect` | Design review or bootstrap constraints |\n| `/ddw:upgrade` | Upgrade project to latest plugin version |\n| `/ddw:auto` | Overnight orchestrator. Loops the pipeline autonomously; logs blockers to morning inbox |\n\nIn the happy path you only ever need: `/ddw:decision` → `/ddw:task` → `/ddw:sendit` → `/ddw:close` (local mode) or `/ddw:decision` → `/ddw:task` → `/ddw:sendit` → `/ddw:pr` → `/ddw:close` (team-PR mode, when `merge.mode: \"github-pr\"`). The worktree creation and teardown happen automatically.\n\n---\n\n## Tips for AI-Powered Dev\n\n- **Run `/clear` after each task**\n  Context builds up and slows things down. Fresh start = fast start.\n\n- **Multitask, no fiddling**\n  `/ddw:sendit` creates an isolated worktree per task automatically — port offset (`.env.ddw`), symlinked `.env*`, fresh `task/TASK-id` branch. `/ddw:close` removes the worktree when the task ships. `maxConcurrent` in `ddw.json` caps how many you can run in parallel.\n\n- **Don't overthink Git early on**\n  Manual is fine. When it gets tedious, let the AI handle it.\n\n- **Stuck? Just type \"next?\"**\n  DDW will scan your pipeline and tell you what to do.\n\n- **AI asks you to verify something?**\n  Check it, then say \"confirmed\" or \"OK\" to move on. (Do actually check though.)\n\n- **Started in English but prefer Japanese?**\n  Just ask \"日本語で進めて\" — it'll switch. Docs too. ([日本語 README](README.ja.md))\n\n---\n\n## Documentation\n\n- [Detailed Guide](GUIDE.md) — full workflow reference, architecture, hook diagrams, agent profiles\n\n## License\n\n[MIT](LICENSE)\n",
  "bytes": 4464,
  "sha": "a55c542cc6f3003b8f6685b5c8beaa5a591759968d7e06aa078777a3367bd9bb",
  "repo_slug": "inalps/ddw",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_inalps_ddw_ddw_07832a43/readme"
}