| name | execute |
| description | Reads PLAN.md and executes each migration step sequentially. Applies transformations file by file, consulting domain-specific reference patterns provided by loaded migration skills. Use after the plan stage has produced PLAN.md.
|
Execute Stage
Executes the approved migration plan from PLAN.md, one file at a time.
Works autonomously — processes all items in sequence without waiting.
References
- Check
/opt/skills/*/references/ for domain-specific transformation patterns
(import maps, annotation replacements, file-type handling rules) from loaded
migration skills
Startup Sequence
- Read
PLAN.md from the repo root — read it ONCE
- Check for reference files in other loaded skills at
/opt/skills/*/references/
for domain-specific transformation patterns
- Begin executing steps in order, starting with Step 1
Do NOT read any source files before starting Step 1.
Domain-specific transformation patterns (import maps, annotation replacements,
file-type handling rules) are provided by migration skills loaded alongside this
stage skill. Consult their reference files for detailed patterns.
Per-File Execution Loop
For each step in PLAN.md, follow this exact sequence:
1. Read the target file
2. Apply transformations per the step's instructions and reference patterns
3. Write the modified file
4. Run: git add -A && git commit -m "<describe the migration change>"
5. Move to the next step immediately
Guardrails
- You MUST attempt every item in PLAN.md in order. Do not skip items.
- After completing each item, note it mentally before moving to the next.
- Do not re-read PLAN.md after every item — read it once, work through the list.
- If you cannot complete an item, note the reason and move to the next.
Do not get stuck on one item.
Important
- Work through ALL items — completeness matters more than perfection
- Do NOT run builds or tests — that is the verify stage's job
- Do NOT modify PLAN.md
- Commit after each step — the harness pushes your commits to the remote
