Executing Plans

When to use batched mode:

• Large plans where you want incremental review
• Plans with uncertain requirements
• When explicitly requested by user

Autonomous is default because:

• Reduces back-and-forth for well-defined plans
• Faster end-to-end execution
• Blockers still trigger stops

Note: This is a default behavior change. Previous behavior was batched. Users expecting checkpoints should add --batched.

The Process

Step 1: Load and Review Plan

1. Read plan files from plans/active/{plan-name}/implementation/
2. Review critically - identify any questions or concerns about the plan
3. If concerns: Raise them with your human partner before starting
4. If no concerns: Create TodoWrite and proceed

Step 2: Execute Tasks

Autonomous Mode (default):

For each task in sequence:

1. Mark as in_progress
2. Follow each step exactly (plan has bite-sized steps)
3. Run verifications as specified
4. Mark as completed
5. Continue to next task (don't stop)

Stop ONLY when:

• All tasks complete
• Hit a blocker (test failure, missing dependency, unclear instruction)
• Verification fails repeatedly

Batched Mode (--batched):

Execute in batches of 3 tasks:

1. Complete 3 tasks following steps above
2. Stop and report progress
3. Wait for feedback
4. Continue with next batch

Step 3: Handle Blockers

When hitting a blocker in either mode:

1. Stop immediately
2. Report what blocked you
3. Report what was completed so far
4. Wait for guidance

Don't guess past blockers - ask for help.

Step 4: Report Completion

Autonomous mode completion report:

## Execution Complete: {plan-name}

### Tasks Completed: {X}/{Y}

### Summary
| Task | Status | Commit |
|------|--------|--------|
| Task 1 description | ✓ Done | abc123 |
| Task 2 description | ✓ Done | def456 |

### Test Results
- Final suite: {X}/{Y} passing

### Blockers Encountered
- [None / List any and how resolved]

### Ready for: implementation-review

Batched mode report: Same as above but for current batch, with "Ready for feedback."

Step 5: Complete Development

After all tasks complete and verified:

• Announce: "I'm using the finishing-a-development-branch skill to complete this work."
• REQUIRED SUB-SKILL: Use gremlins:finishing-a-development-branch
• Follow that skill to verify tests, present options, execute choice

When to Stop and Ask for Help

STOP executing immediately when:

• Hit a blocker mid-batch (missing dependency, test fails, instruction unclear)
• Plan has critical gaps preventing starting
• You don't understand an instruction
• Verification fails repeatedly

Ask for clarification rather than guessing.

When to Revisit Earlier Steps

Return to Review (Step 1) when:

• Partner updates the plan based on your feedback
• Fundamental approach needs rethinking

Don't force through blockers - stop and ask.

Remember

• Review plan critically first
• Autonomous is default - run to completion unless blocked
• Batched mode - stop every 3 tasks for checkpoint
• Follow plan steps exactly
• Don't skip verifications
• Reference skills when plan says to
• Stop when blocked, don't guess