| name | pwrl-work |
| description | Execute implementation work efficiently through 4-phase micro-skill pipeline |
| argument-hint | [Task file, plan doc path, or work description. Leave blank to use latest plan/task] |
PWRL Work
Execute implementation work through a deterministic 4-phase pipeline: triage input, prepare environment, implement with test-first discipline, and review code quality.
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 work would you like to execute? Provide a task file path, plan file, or describe the work to be done."
- Provide clear recovery suggestions when errors occur
Purpose
Transform task files, plans, or prompts into completed working code through systematic execution:
- Triage Input — Classify and validate input (task file, plan file, bare prompt, or latest task)
- Prepare Environment — Repository verification, ambiguity resolution, branch setup, verification commands
- Execute Implementation — Test-first implementation with incremental verification and quality gates
- Review & Verify — Code review, scope check, diff quality, documentation
Usage
/pwrl-work
/pwrl-work docs/tasks/to-do/2026-06-11-U2-email-validation.md
/pwrl-work docs/plans/2026-06-11-003-skill-architecture.md
/pwrl-work "add email validation to user signup"
Architecture
Direct sequence of 4 micro-skills with deterministic artifact flow:
Input
↓
Phase 0: pwrl-work-triage
├ Input: task file, plan file, bare prompt, or empty
├ Output: triage artifact (unit_id, files, acceptance_criteria, dependencies)
↓
Phase 1: pwrl-work-prepare
├ Input: triage artifact
├ Processing: repo verification, ambiguity resolution, branch strategy, verification commands
├ Output: prepare artifact (branch, verification_commands, environment state)
↓
Phase 2: pwrl-work-execute
├ Input: prepare artifact
├ Processing: scaffolding, test-first implementation, quality gates
├ Output: execute artifact (files changed, tests passing, build/lint status)
↓
Phase 3: pwrl-work-review
├ Input: execute artifact
├ Processing: scope check, diff review, test review, documentation check
├ Output: review artifact (approval status, ready_to_ship)
↓
COMPLETE
Each phase produces an explicit artifact (YAML frontmatter + structured data) consumed by the next phase. Enables resumability, traceability, and independent testing.
Phase Summary
Phase 0: Triage — Validate input, classify work, extract requirements, set interaction mode. See pwrl-work-triage/SKILL.md.
Phase 1: Prepare — Verify repository, resolve ambiguities, setup branch and environment. See pwrl-work-prepare/SKILL.md.
Phase 2: Execute — Implement with test-first discipline, verify quality gates. See pwrl-work-execute/SKILL.md.
Phase 3: Review — Verify scope, review code quality, get approval. See pwrl-work-review/SKILL.md.
Interaction Mode Propagation
Interaction mode (detailed | smart | yolo) is set in Phase 0 (via pwrl-work-triage Step 5) and read at the start of each subsequent phase. The mode is stored in the triage artifact's interactionMode field and propagated through every downstream phase. Determines whether confirmation steps execute or are skipped.
detailed — Pause at every phase transition (Triage → Prepare → Execute → Review); show generated artifacts; require explicit approval to proceed. Best for complex work, unfamiliar codebases, and learning.
smart — Phases run automatically; pause only when the next phase produces a HIGH-risk operation (destructive git, irreversible API calls, schema-breaking migrations). v1 simplification: behaves like Yolo with a single confirmation prompt at workflow start. See docs/learnings/pattern/interaction-mode-three-mode-propagation-2026-06-29.md §"Future Refinements" for the full risk-classification roadmap.
yolo — Every phase runs automatically; only the final review report is shown. Fastest. Best for straightforward, well-understood work and time-sensitive hotfixes.
Exception: Error recovery steps always pause the pipeline, regardless of mode. Downstream phases that still assume a legacy two-value enum must treat any value other than detailed as yolo until upgraded; smart is new as of 2026-06-29.
Quality Criteria
- Requirements from task/plan fully implemented
- Behavior changes covered by tests
- Tests pass after each logical unit
- Code follows project patterns and conventions
- No debug code, console.logs, or commented code left behind
- Scope stayed tight (no unrelated changes)
Rules
- Clarify ambiguities upfront — Don't proceed if approach is unclear
- Verify incrementally — Run checks frequently, catch issues early
- Test-first discipline — Write tests before implementing
- One scope at a time — Complete and review each unit separately
- No scope creep — Get approval before expanding beyond task
- Quality gates — All checks must pass before marking ready
Error Handling & Recovery
For comprehensive error scenarios, recovery strategies, and escalation rules, see references/error-handling.md.
Support Files
pwrl-work-triage/references/triage-input-protocol.md — Phase 0 specification
pwrl-work-prepare/references/prepare-environment-protocol.md — Phase 1 specification
pwrl-work-execute/references/execute-implementation-protocol.md — Phase 2 specification
pwrl-work-review/references/review-quality-protocol.md — Phase 3 specification
When to Use
- Use when executing a clear task from a task file
- Use when extracting and executing units from a plan
- Use for well-defined work with clear acceptance criteria
- For exploratory work without clear requirements, use
/pwrl-plan first