Agent-Skills.md

Agent Skills: Executing Plans

Use when given a complete implementation plan to execute in controlled batches with review checkpoints - loads plan, reviews critically, executes tasks in batches, reports for review between batches

UncategorizedID: timbuchinger/loadout/executing-plans

Repository

timbuchinger/loadout
timbuchinger

Install this agent skill to your local

pnpm dlx add-skill https://github.com/timbuchinger/loadout/tree/HEAD/skills/executing-plans

Skill Files

Browse the full folder contents for executing-plans.

Download Skill

Loading file tree…

skills/executing-plans/SKILL.md

Skill Metadata

Name
executing-plans
Description
Use when given a complete implementation plan to execute in controlled batches with review checkpoints - loads plan, reviews critically, executes tasks in batches, reports for review between batches

Executing Plans

Overview

Load plan, review critically, execute tasks in batches, report for review between batches.

Core principle: Batch execution with checkpoints for review and feedback.

When to Use

  • Partner provides a complete implementation plan
  • Plan has detailed tasks with verification steps
  • Work benefits from checkpoint reviews
  • Executing work created by writing-plans skill

The Process

Step 1: Load and Review Plan

  1. Read the plan file (typically in docs/plans/)
  2. Review critically - identify any questions or concerns
  3. Check for:
    • Missing information or unclear steps
    • Potential blockers or dependencies
    • Inconsistencies or gaps in approach
  4. If concerns: Raise them with your partner before starting
  5. If no concerns: Create todo list and proceed

Step 2: Execute Batch

Default: First 3 tasks

For each task in the batch:

  1. Mark task as in_progress in todo list
  2. Follow each step exactly (plans have bite-sized steps)
  3. Run all verification steps as specified
  4. Confirm tests pass and changes work
  5. Mark task as completed

Follow the plan precisely:

  • Execute steps in order
  • Run commands as specified
  • Verify expected output
  • Don't skip verification steps
  • Don't deviate from plan unless blocked

Step 3: Report

When batch complete:

  • Summarize what was implemented
  • Show verification output (test results, command output)
  • Highlight any issues or deviations
  • Say: "Ready for feedback."

Keep reports concise but complete:

  • Which tasks were completed
  • Key changes made
  • Verification results
  • Any concerns or questions

Step 4: Continue

Based on feedback:

  • Apply requested changes if needed
  • Execute next batch (next 3 tasks)
  • Repeat until all tasks complete

Step 5: Complete Development

After all tasks complete and verified:

  • Run full test suite
  • Verify all changes work together
  • Review what was built
  • Consider next steps (PR, deployment, etc.)

When to Stop and Ask

STOP executing immediately when:

  • Hit a blocker mid-batch (missing dependency, test fails unexpectedly)
  • Instructions are unclear or ambiguous
  • Verification fails repeatedly
  • Discover plan has critical gaps
  • Approach seems incorrect

Ask for clarification rather than guessing.

Don't try to work around blockers - stop and get guidance.

When to Revisit Earlier Steps

Return to Review (Step 1) when:

  • Partner updates the plan based on feedback
  • Fundamental approach needs rethinking
  • New information changes requirements

Don't force through blockers - stop and ask.

Best Practices

Be systematic:

  • Follow plan order unless advised otherwise
  • Complete one task fully before starting next
  • Don't skip ahead or batch tasks differently

Be thorough:

  • Run all verification steps
  • Check for unexpected side effects
  • Verify changes work as intended

Be communicative:

  • Report clearly at checkpoints
  • Flag concerns early
  • Ask when uncertain

Be precise:

  • Follow instructions exactly
  • Use exact file paths from plan
  • Run exact commands specified

Example Execution

Batch 1 Report

Completed tasks 1-3:

Task 1: Email Validation
- Created src/validators/email.py
- Created tests/validators/test_email.py
- Tests pass: 2/2 ✓

Task 2: User Model Update
- Modified src/models/user.py
- Added email field validation
- Tests pass: 5/5 ✓

Task 3: API Endpoint
- Created src/api/auth.py
- Integrated email validation
- Tests pass: 3/3 ✓

All verifications passed. Ready for feedback.

Common Mistakes to Avoid

  • Don't deviate from the plan without approval
  • Don't skip verification steps to save time
  • Don't batch more tasks than specified
  • Don't continue when blocked
  • Don't assume what the plan meant - ask

Quick Reference

| Step | Action | When to Stop | |----------|-------------------------------|----------------| | 1. Load | Review plan critically | Concerns found | | 2. Execute | Implement 3 tasks | Blocker hit | | 3. Report | Show results, ask for feedback | - | | 4. Continue | Next batch based on feedback | All tasks done | | 5. Complete | Final verification | - |

Final Rule

Follow the plan exactly.
Stop when blocked or uncertain.
Report at checkpoints.

Trust the plan, but verify everything.