tasks

tasks

Extract a structured tasks.md from an existing plan.md. Pure extraction — no delegation. Produces checkbox tasks with parallel markers, story labels, and phase checkpoints.

Prereqs

• docs/maxi/constitution.md must exist — hard stop if missing
• Spec status must be planned. Other statuses:
  • drafting / specified / clarified: stop — "Spec must reach planned status first. Run /maxi:plan."
  • tasked or later: stop — "Tasks already extracted. Proceed to /maxi:analyze or /maxi:implement."

Process

1. Read inputs — load plan.md as primary source; also load research.md, data-model.md, contracts/ if they exist alongside it in docs/maxi/specs/NNNN-slug/
2. Map tasks to user stories — for each user story in spec.md, collect the tasks from plan.md that implement it. Tag each task with [US1], [US2], etc.
3. Identify parallel tasks — mark with [P] any task that touches different files from all other tasks in the same phase (no shared-file writes, no dependency on concurrent tasks)
4. Assign sequential IDs — number all tasks T001, T002, ... in phase order. No letters, no "Task N", no "Step N".
5. Structure into phases — verify tasks-template.md exists (Read tool) before proceeding; if missing, stop: "Cannot proceed — tasks-template.md is missing. Please reinstall the maxi plugin." Then follow the template:
   • Phase 1: Setup (project init, no deps)
   • Phase 2: Foundational (blocks all user stories — CRITICAL warning)
   • Phase 3+: One phase per user story
   • Final phase: Polish & Cross-Cutting Concerns
   • A Checkpoint line after each phase
6. Write tasks.md — output to docs/maxi/specs/NNNN-slug/tasks.md following the template schema. Include Dependencies & Execution Order section.
7. Transition status — update spec.md frontmatter status: planned → tasked; also set updated: [today's ISO date] on spec.md and on tasks.md. When creating tasks.md, set its frontmatter: slug and spec_slug from spec, created and updated to today's ISO date.
8. Report — "Tasks written to docs/maxi/specs/NNNN-slug/tasks.md (status: tasked). Next: /maxi:analyze (recommended) or /maxi:implement."

Task Format

- [ ] T001 [P] [US1] Create [Entity] model in src/models/entity.py
- [ ] T002 [US1] Implement [Service] in src/services/service.py (depends on T001)
- [ ] T003 [P] [US2] Create [OtherEntity] model in src/models/other.py

Rules:

• [P] = safe to run in parallel (different files, no dependency)
• [USN] = which user story (must match a story from spec.md)
• No [P] on tasks that share files or have dependencies within the same phase
• Include exact file paths in descriptions

Phase Structure

## Phase 1: Setup

- [ ] T001 ...
- [ ] T002 [P] ...

**Checkpoint**: Setup complete — foundational phase can begin.

---

## Phase 2: Foundational ⚠️ BLOCKS ALL USER STORIES

- [ ] T003 ...

**Checkpoint**: Foundation ready — user story phases can begin.

---

## Phase 3: User Story 1 - [Title] (Priority: P1) 🎯 MVP

**Goal**: ...
**Independent Test**: ...

- [ ] T010 [P] [US1] ...
- [ ] T011 [US1] ...

**Checkpoint**: User Story 1 complete and independently testable.

Artifact reference links

When this skill emits prose that references another maxi artifact (an ADR, spec, plan, tasks, constitution, or repo file) — in an artifact body or in a chat report — render it as a relative Markdown link, not a bare slug/number/code span:

• Visible text = the target filename without .md (an ADR slug like 0003-constitution-decoupled-from-claudemd; for generic spec artifacts use <feature-dir>/<name>, e.g. 0002-migrate-adr-review-fixes/spec; non-.md files keep their full name).
• URL = a relative path from the referencing file's directory (workspace-root-relative for chat reports).
• Do NOT link frontmatter data values (related_adrs entries stay bare slugs) or within-document IDs (FR-012, section names).
• Applies forward-only — do not retro-edit existing artifacts.

Critical Rules

• Read plan.md before writing anything. All tasks must be traceable to plan.md content.
• Pure extraction only. Do NOT invent tasks that are not in plan.md. "Being thorough" is not a justification for adding tasks. Every task must trace to a step in plan.md.
• Apply [P] and [USN] markers even if plan.md already has checkboxes. If plan.md has flat task items, you MUST restructure them: add phase grouping, add [USN] labels, identify [P] markers. Copying plan.md verbatim is NOT acceptable — it will be missing structure.
• Every task needs a story label. If a task doesn't belong to a user story (Setup/Foundational/Polish), omit [USN] label. All implementation tasks for a user story MUST have one.
• IDs are sequential and numeric. T001–T999. Never T1, Task 1, Step A.
• Parallel markers require no shared files. If two tasks touch the same file, neither gets [P].
• MVP Checkpoint after Phase 3 (US1). Always mark the P1 user story checkpoint as the MVP.
• Phase structure is required even for single-story features. One user story = at minimum: Setup → Foundational → Phase 3 (US1) → Polish. A flat list with no phases is never acceptable.
• Template schema required. Follow tasks-template.md structure. Include Dependencies section.

Red Flags

• Flat list of tasks with no phases → structure into phases even for single-story features
• No [P] markers anywhere → identify parallel opportunities
• No [USN] labels → map every implementation task to a story
• IDs like "Task 1", "Step A", or "T1" → use T001 format
• No Checkpoint lines → add between every phase
• tasks.md written to wrong directory → only docs/maxi/specs/NNNN-slug/tasks.md
• tasks.md looks identical to plan.md (no added markers or structure) → extraction not done
• Extra tasks not found in plan.md → remove invented tasks, only extract what plan.md contains

Rationalization Counters

Rationalization	Counter
"plan.md already has checkbox tasks, I'll just copy them"	plan.md tasks are unstructured. You MUST add [P] markers, [USN] labels, phase grouping, and Checkpoints. Copy is never acceptable.
"I'll add a few extra tasks to be thorough"	Pure extraction only. Every task traces to plan.md. No invented tasks.
"This feature only has one user story, phases are overkill"	Phase structure is required regardless of story count: Setup → Foundational → US1 phase → Polish.
"I don't see a foundational phase in plan.md"	Every tasks.md has a Phase 2: Foundational, even if minimal (e.g., one task: "Configure project dependencies").
"Tasks marked [P] in plan.md might have hidden dependencies"	Re-evaluate based on file-level analysis. If plan.md marks something parallel but they share a file, remove the [P] marker.