🧭Subagents
Long migration planner
Uses the Plan subagent to survey current usage and propose three migration strategies without polluting main context.
- Surfaces
- skill · subagent
- Complexity
- advanced
- Trigger
- natural
- Est. tokens
- 9,500
What It Does
Delegates the exploration phase of a big migration (framework upgrade, language switch, DB swap) to the Plan subagent. Returns three strategies — big bang, strangler fig, parallel implementations — each with risk, timeline, and rollback notes.
When It Triggers
- ▸"Plan the migration from Flask to FastAPI"
- ▸"How do we upgrade React 17 → 19?"
- ▸"We need to move off MongoDB — plan it"
SKILL.md
markdown
---
name: long-migration-planner
description: Plans multi-sprint migrations (framework upgrades, DB migrations, language rewrites) using the Plan subagent to explore options without polluting the main context. Use when user says "plan this migration", "upgrade from X to Y", or "how do we migrate".
---
# Long Migration Planner
## Steps
1. Delegate to Plan subagent: "Survey current usage of [library/pattern]"
2. Ask Plan subagent to produce 3 migration strategies:
- Big bang · Strangler fig · Parallel implementations
3. For each strategy: risk level, estimated weeks, rollback path
4. Main agent synthesizes: recommended path + phase-by-phase checklist
5. Save the plan to `docs/migrations/<target>-migration.md`
Why Plan Subagent
Surveying current usage can touch hundreds of files. Keeping that exploration inside a Plan subagent means the main context stays focused on the strategic decision, not the raw survey data.
Gotchas
- ▸The Plan subagent cannot execute destructive actions — it only surveys.
- ▸Save the final plan to disk so it survives across sessions.
- ▸Revisit the plan after phase 1 — real data from phase 1 usually invalidates assumptions in phase 3+.