Execute

Executing Implementation Plans

Overview

Load plan, review critically, execute tasks in batches, report for review between batches. Stop immediately when blocked - never guess.

Core principle: Batch execution with checkpoints for review.

Announce at start: "I'm using the execute skill to implement this plan."

The Process

Step 1: Load and Review Plan

1. Read the plan file completely
2. Review critically - identify any questions or concerns
3. If concerns: Raise them before starting. Don't proceed with unclear instructions.
4. If no concerns: Create TodoWrite with all tasks, proceed to Step 2

Questions to ask:

• Are file paths clear and complete?
• Is the code complete or does it have placeholders?
• Do I understand the test expectations?
• Are there implicit dependencies not listed?

Step 2: Execute Batch

Default batch size: 3 tasks

For each task in the batch:

1. Mark task as in_progress in TodoWrite
2. Follow each step exactly as written
3. Run verifications as specified in plan
4. Mark as completed only when ALL steps pass
5. Capture learnings: kodo reflect --signal "Pattern that worked: ..."

Execute steps literally:

• If plan says "run cargo test" - run exactly that
• If plan says "expected output: FAIL" - verify you see FAIL
• If plan shows code - use that exact code

Step 3: Report

When batch complete:

Completed tasks 1-3:
- Task 1: [summary] - PASS
- Task 2: [summary] - PASS
- Task 3: [summary] - PASS

Verification output:
[key test results]

Ready for feedback before continuing.

Wait for user response before proceeding.

Step 4: Continue or Adjust

Based on feedback:

• "Continue" - Execute next batch
• "Fix X" - Apply changes, re-verify, then continue
• "Stop" - Save progress, note remaining tasks

Repeat Steps 2-4 until complete.

Step 5: Completion

After all tasks complete:

1. Run full test suite
2. Verify all commits are clean
3. Offer code review: "Ready for review with kodo:review skill?"
4. Capture session learnings: kodo reflect

When to Stop Immediately

STOP executing when:

• Hit a blocker (missing dependency, unclear instruction)
• Test fails unexpectedly (not expected failure)
• Plan has critical gaps
• You don't understand what a step means
• Verification fails after 2 attempts

Ask for clarification rather than guessing.

BLOCKED on Task 3, Step 2:

Plan says: "Run migration"
Problem: No migration file exists at specified path

Options:
A) Create migration file (show me what to create)
B) Skip this step (may cause issues in Task 4)
C) Abort and revise plan

Which approach?

Progress Tracking

Use TodoWrite to track:

- [x] Task 1: Add config struct
- [x] Task 2: Implement parser
- [ ] Task 3: Add validation (IN PROGRESS)
- [ ] Task 4: Wire up CLI
- [ ] Task 5: Add tests

Update after each task completion, not at batch boundaries.

Integration with Kodo

During execution:

kodo reflect --signal "This pattern worked well"
kodo query "error handling"  # When stuck, check context

After completion:

kodo reflect              # Capture all session learnings
kodo track link #123      # Update linked GitHub issue

Key Principles

• Review plan critically first - Don't blindly execute broken plans
• Follow steps exactly - Plans are pre-validated, trust them
• Don't skip verifications - They catch issues early
• Stop when blocked - Guessing creates more problems
• Report and wait - User feedback between batches

Red Flags

You're doing it wrong if:

• Executing without reviewing plan first
• Skipping verification steps
• Continuing past unexpected failures
• Batches larger than 3 tasks without checkpoint
• Guessing when instructions are unclear
• Not updating TodoWrite progress