OpenParable
Skip to content
← Back to skills

Phase 4 Execute

Productivity
phase-4-execute

Phase 4 Execute is an Agent Skill for productivity published in the majiayu000/claude-skill-registry repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Execute phase skill for plan management. DUMB TASK RUNNER that executes tasks from TASK-*.toon files sequentially for execute and finalize phases.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/majiayu000/claude-skill-registry/blob/HEAD/skills/workflow/phase-4-execute-cuioss-plan-marshall/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/phase-4-execute, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Phase Execute Skill

Role: DUMB TASK RUNNER that executes tasks from TASK-*.toon files sequentially.

Execution Pattern: Locate current task → Execute steps → Mark progress → Next task

Phases Handled: execute, finalize

CRITICAL: Use manage-* scripts via Bash for plan file updates (Edit/Write tools trigger permission prompts on .plan/ directories).


Scripts

ScriptPurpose
pm-workflow:manage-config:manage-configConfig field access
pm-workflow:manage-lifecycle:manage-lifecyclePhase routing and transitions
plan-marshall:logging:manage-logWork log entries
pm-workflow:manage-tasks:manage-tasksTask and step management
pm-workflow:manage-references:manage-referencesReference file CRUD

Standards (Load On-Demand)

Workflow

Read standards/workflow.md

Contains: Task execution pattern, phase transition, auto-continue behavior

Operations

Read standards/operations.md

Contains: Delegation patterns for builds, quality checks, PR creation

Finalize Configuration (from config.toon)

For finalize phase, read finalize configuration directly from config.toon:

python3 .plan/execute-script.py pm-workflow:manage-config:manage-config get-multi \
  --plan-id {plan_id} \
  --fields create_pr,verification_required,verification_command,branch_strategy

Returns only the required finalize fields in a single call: create_pr, verification_required, verification_command, branch_strategy.

These fields are written during init based on domain configuration.


Execution Loop

Step 0: Get Routing Context (Once at start)

Get current phase, skill routing, and progress in a single call:

python3 .plan/execute-script.py pm-workflow:manage-lifecycle:manage-lifecycle get-routing-context \
  --plan-id {plan_id}

Returns:

status: success
plan_id: {plan_id}
current_phase: execute
skill: pm-workflow:phase-4-execute
skill_description: Execute phase skill for task implementation
total_phases: 4
completed_phases: 2
phases:
- init: complete
- refine: complete
- execute: in_progress
- finalize: pending

Use current_phase for logging, skill for dynamic routing, and completed_phases/total_phases for progress display.

Step 0.5: Log Phase Start (Once per phase)

At the start of execute or finalize phase:

python3 .plan/execute-script.py plan-marshall:logging:manage-log \
  work {plan_id} INFO "[STATUS] (pm-workflow:phase-4-execute) Starting {phase} phase"

For each task in current phase:

Step 1: Locate Task with Context

python3 .plan/execute-script.py pm-workflow:manage-tasks:manage-tasks next \
  --plan-id {plan_id} \
  --include-context

Returns next task with status pending or in_progress, including embedded goal context (title, body) for immediate use without additional script calls.

Step 2: Execute Steps

For each step in task's steps[] array:

  1. Parse the step text
  2. Execute the action (delegate if specified)
  3. Mark step complete via manage-tasks:step-done

Step 3: Mark Step Complete

python3 .plan/execute-script.py pm-workflow:manage-tasks:manage-tasks step-done \
  --plan-id {plan_id} \
  --task {task_number} \
  --step {step_number}

Step 3.5: Log Task Completion

After each task completes:

python3 .plan/execute-script.py plan-marshall:logging:manage-log \
  work {plan_id} INFO "[OUTCOME] (pm-workflow:phase-4-execute) Completed {task_id}: {task_title} ({steps_completed} steps)"

Step 4: Next Task or Phase

  • If more tasks in phase → Continue to next task
  • If phase complete → Log phase outcome and auto-transition to next phase
  • If all phases complete → Mark plan complete

Step 5: Log Phase Completion (When phase completes)

python3 .plan/execute-script.py plan-marshall:logging:manage-log \
  work {plan_id} INFO "[STATUS] (pm-workflow:phase-4-execute) Completed {phase} phase: {tasks_completed} tasks"

Delegation

When checklist items specify delegation, invoke the appropriate agent/skill:

Checklist PatternDelegation
"Run build" / "maven" / "npm"See standards/operations.md
"Delegate to {agent}"Task: {agent}
"Load skill: {skill}"Skill: {skill}
"Run /command"SlashCommand: /command

Auto-Continue Behavior

Execute continuously without user prompts except:

  • Error blocks progress
  • Decision genuinely required
  • User explicitly requested confirmation

Do NOT prompt for:

  • Phase transitions
  • Task transitions
  • Routine confirmations

Phase Transition

When transitioning from execute phase to finalize:

python3 .plan/execute-script.py pm-workflow:manage-lifecycle:manage-lifecycle transition \
  --plan-id {plan_id} \
  --completed execute

This automatically updates status.toon and moves to the next phase.


Error Handling

On any error, first log the error to work-log:

python3 .plan/execute-script.py plan-marshall:logging:manage-log \
  work {plan_id} ERROR "[ERROR] (pm-workflow:phase-4-execute) {task_id} failed - {error_type}: {error_context}"

Script Failure (Lessons-Learned Capture)

ON SCRIPT FAILURE: When any script execution fails (exit != 0):

  1. Log error to work-log (see above)
  2. Capture error context (script path, exit code, stderr)
  3. Continue with normal error recovery (retry, fail task, etc.)

Other Errors

ErrorOptions
Build failureFix and retry / View log / Skip task
Test failureFix tests / View details / Skip task
Dependency not metComplete dependency / Skip check

Integration

Command Integration

  • /plan-execute - Primary command invoking this skill

Skills Used

SkillCommandPurpose
pm-workflow:manage-lifecycleget-routing-contextPhase, skill routing, progress
pm-workflow:manage-configget-multiFinalize config fields
pm-workflow:manage-tasksnext --include-contextTask with goal context
plan-marshall:logging:manage-logworkWork log entries
pm-workflow:manage-references-Reference file CRUD
pm-workflow:workflow-integration-git-Commit operations

Related Skills

  • plan-init - Creates plan structure (request.md, config, status)
  • solution-outline - Creates deliverables from request
  • task-plan - Creates tasks from deliverables