Execution Plans Skill

Treat execution plans as first-class artifacts, versioned in the repository.

Directory Structure

text
1.agents/
2├── plans/
3│   ├── active/        # Plans currently in progress
4│   ├── completed/     # Successfully finished plans
5│   └── abandoned/     # Plans that were stopped (with rationale)
6└── debt/
7    └── tech-debt-registry.md  # Known technical debt items

Triggers

Plan Template

Use .agents/plans/TEMPLATE.md as the starting point for new plans.

Required Sections

Workflow

Creating a Plan

1. Copy TEMPLATE.md to .agents/plans/active/{slug}.md
2. Fill metadata (status: In Progress, created: today, owner: agent name)
3. Define objectives as checkboxes
4. Link to related issue/PR

Updating Progress

1. Check off completed objectives
2. Add entry to Progress Log table with date, update, and agent name
3. Update blockers if any emerge

Logging Decisions

Add row to Decision Log table:

Completing a Plan

1. Verify all objectives checked
2. Update status to Completed
3. Move file from active/ to completed/
4. Add final progress entry

Blocking a Plan

1. Update status to Blocked
2. Document impediment in Blockers section
3. Add progress entry noting the block

Abandoning a Plan

1. Update status to Abandoned
2. Document rationale in blockers or final progress entry
3. Move file from active/ to abandoned/

Integration

• Session logs reference active plans when relevant
• Planner skill creates plans here when executing complex work
• Retrospectives link back to completed plans

Anti-Patterns

Verification

After creating a plan:

• File in .agents/plans/active/
• Metadata section complete
• At least one objective defined
• Linked to issue or PR

After completing:

• All objectives checked
• Final progress entry added
• File moved to completed/