| name | pwrl-tasks |
| description | Slice implementation plans into granular, executable task files. Use when: breaking down plans from pwrl-plan; creating actionable tasks for pwrl-work; need detailed task files in docs/tasks/to-do. |
| argument-hint | [Path to plan file, or leave blank to find latest plan] |
PWRL Tasks
Slice a plan into granular, executable task files for /pwrl-work.
Interaction Method
- Use platform's
ask_user_question, ask_user, ask_user_input, vscode/askQuestions or any available extension/tool for user interaction for all decisions
- Ask one question at a time
- Use multiple-choice questions when possible
- If input is empty, ask: "What plan would you like to slice into tasks? Provide a plan file path or leave blank to auto-discover the latest plan."
- Provide clear recovery suggestions when errors occur
Purpose
- Create ready-to-execute work items in
docs/tasks/to-do/
- Preserve unit IDs and dependencies from the plan
- Enrich tasks with implementation steps, tests, and acceptance criteria
Usage
/pwrl-tasks
/pwrl-tasks docs/plans/2026-05-01-001-auth.md
Input
Plan document path, or blank to auto-discover the latest plan in docs/plans/.
Support Files
references/task-template.md — Canonical task structure (body + frontmatter)
references/index-template.md — Canonical docs/tasks/INDEX.md structure
references/dependency-resolution.md — Dependency modeling and validation guidance
references/examples.md — Examples of good tasks and indexes
references/github-issues-integration.md — Optional GitHub Issues sync rules
Workflow
Phase 0: Select Interaction Mode
Before locating the plan, ask the user to choose their engagement level for this slicing workflow. Use the platform's ask_user_question extension (or equivalent) to present the following three options:
Question: "How would you like to proceed with this task slicing?"
Options:
- Detailed (Step-by-Step) — Show each generated task file to the user before writing; pause for ambiguous dependency resolutions; require explicit approval before updating the index. Maximum control. Best for complex plans where dependency resolution is non-obvious.
- Smart (Risk-gated automation) — Write all task files automatically; pause only for ambiguous dependency resolutions (e.g., a unit that references a missing unit-id) and for the index update. v1 simplification: behaves like Yolo with a single confirmation prompt at workflow start.
- Yolo (Full Automation) — Write all task files automatically; auto-resolve dependencies with conservative defaults (e.g., assume a missing dependency is a future unit, do not block on it); auto-update the index. Fastest. Best for routine plans and trusted plan templates.
Store the selection in the orchestrator context for use by Phases 2 and 3:
interactionMode: detailed | smart | yolo
Note: A future enhancement could read the mode from .pwrlrc.json so users who always want Yolo don't get re-prompted. Out of scope for plan 2026-06-29-001 — see docs/learnings/pattern/interaction-mode-three-mode-propagation-2026-06-29.md §"Future Refinements".
Interaction Mode Propagation
The interactionMode value is set in Phase 0 and consumed by Phases 2 and 3:
- Phase 2 (Generate Task Files):
- Detailed: Show each generated task file to the user before writing; pause for ambiguous dependency resolutions.
- Smart: Write all task files automatically; pause only for ambiguous dependency resolutions.
- Yolo: Write all task files automatically; auto-resolve dependencies with conservative defaults.
- Phase 3 (Generate Index and Report):
- Detailed: Show the proposed index update to the user before writing; require explicit approval.
- Smart: Write the index automatically; report the diff afterward.
- Yolo: Write the index automatically; no diff preview.
Phase 1: Locate and Read the Plan
- If a path is provided, read that plan; otherwise scan
docs/plans/ for the latest active/draft plan.
- If no plan exists, ask the user to create one with
/pwrl-plan or provide a path.
- Extract implementation units (
U1, U2, ...) plus any dependencies and test scenarios.
- Read
docs/learnings/INDEX.md and map relevant learnings to each unit (or mark a learning gap).
Phase 2: Generate Task Files
-
For each unit, create docs/tasks/to-do/YYYY-MM-DD-uX-<slug>.md.
-
Use references/task-template.md as the canonical task structure and keep tasks self-contained.
-
Ensure task frontmatter includes (at minimum):
---
unit-id: U1
plan: docs/plans/YYYY-MM-DD-NNN-name.md
status: to-do
dependencies: [U2]
files: [path/to/file.ts]
learnings: [docs/learnings/pattern/example.md]
---
-
Ensure task body includes: goal, context, steps, edge cases, testing, acceptance criteria, and references.
-
If a unit has no relevant learning, add an explicit step to document one via /pwrl-learnings.
Phase 3: Generate Index and Report
- Create/update
docs/tasks/INDEX.md using references/index-template.md.
- Validate dependencies (see
references/dependency-resolution.md for guidance).
- Report created tasks, critical path, and recommended starting tasks.
Output
- Task files created under
docs/tasks/to-do/
- A task index at
docs/tasks/INDEX.md reflecting statuses and dependencies
Quality Gate Validation: Run /pwrl-phase-checkpoint tasks 3 [artifact-path] to validate all tasks generated and dependencies resolved. See pwrl-phase-checkpoint for validation rules.
Best Practices
- Keep each task independently executable (with dependencies noted)
- Preserve stable unit IDs from the plan; never renumber
- Prefer linking to existing codebase patterns over inventing new conventions
- If GitHub Issues sync is enabled, follow
references/github-issues-integration.md
Rules
- Never delete or overwrite existing tasks without explicit user confirmation.
- Use repository-relative paths only in tasks and indexes.
- If dependency information is ambiguous, ask before encoding it into tasks.