zeus-plan

---

name: zeus-plan description: Convert a spec document into structured prd.json user stories and task.json execution units. Called automatically by zeus-brainstorm, but can also be invoked directly. Supports --version flag to target v2/v3/v4 folders. argument-hint: [--spec ] [--version vN] shell: bash

Zeus Plan - Spec to Executable Artifacts

Convert an approved spec into version-scoped stories and tasks with deterministic IDs and dependency waves.

Preconditions

• Target spec exists and is approved.
• .zeus/{version}/config.json exists.
• .zeus/{version}/prd.json, task.json, and roadmap.json are writable.

Inputs

• --spec <filename> (optional, default latest spec)
• --version vN (optional, default main)

Deterministic workflow

1) Load context and reserve IDs

Read:

• selected spec
• version config
• existing prd.json
• existing task.json

Reserve next IDs:

• US-{NNN} from prd.json
• T-{NNN} from task.json
• M-{NNN} from roadmap.json milestones

2) Create user stories

Extract from acceptance criteria and role/goal language in spec.

Required story shape:

{
  "id": "US-001",
  "title": "One-line story title",
  "description": "As a <role>, I need <capability> so that <outcome>",
  "acceptance_criteria": ["Given/When/Then style criterion"],
  "priority": "high",
  "version_scope": "main",
  "passes": false,
  "created_from_spec": "2026-04-07-auth-design.md",
  "created_at": "2026-04-07T00:00:00.000Z"
}

Priority policy:

• high: direct north star impact
• medium: critical support path
• low: optimization or edge behavior

3) Create executable tasks

Generate small, independently executable units.

Task types:

• infra
• api
• frontend
• docs

Task schema:

{
  "id": "T-001",
  "story_id": "US-001",
  "type": "api",
  "title": "Implement POST /api/auth/login",
  "description": "Include validation, error mapping, and audit logging.",
  "files": ["src/app/api/auth/login/route.ts"],
  "depends_on": [],
  "refactor_of": null,
  "depends_on_existing": [],
  "passes": false,
  "commit_sha": null,
  "ai_log_ref": null,
  "wave" 
   

Brownfield policy:

• If task modifies existing modules, fill refactor_of with a stable module path or identifier.
• If task depends on existing code not represented by Zeus task IDs, list those in depends_on_existing.
• Keep both fields optional for backward compatibility.

4) Compute wave plan

Build dependency DAG from depends_on and assign wave numbers:

• wave 1: no dependencies
• wave N: depends only on lower waves

Reject cyclic dependencies and force correction before write.

5) Update roadmap milestone

Add or update milestone with:

• title from spec
• spec_ref
• generated story/task ID lists
• status: planned

6) Persist artifacts

Write updated:

• .zeus/{version}/prd.json
• .zeus/{version}/task.json
• .zeus/{version}/roadmap.json

7) Commit planning artifacts

git add .zeus/{version}/prd.json .zeus/{version}/task.json .zeus/{version}/roadmap.json
git commit -m "docs(zeus): update prd and task from {spec-name}"

8) Write AI log

Create .zeus/{version}/ai-logs/{ISO-ts}-plan.md with the 3-section contract.

9) Report summary

Return:

• story count,
• task count,
• wave distribution,
• recommended next step: /zeus:execute.

10) Offer test generation

After printing the summary, prompt the user:

Generate test flows now? [yes/no]
(Platforms: android, chrome, ios — runs /zeus:test-gen automatically)

• If yes: invoke /zeus:test-gen --version {version} immediately.
• If no: remind the user they can run /zeus:test-gen at any time before executing.

Quality gates

• no duplicate story/task IDs,
• no empty acceptance criteria,
• no task without a valid story link,
• no wave assignment with unresolved dependency.

Agent collaboration

Use zeus-planner for decomposition quality checks and dependency sanity verification.