Migration Plan Skill: Reversible Steps, Ordered By Risk
Produces an ordered, reversible upgrade plan with a rollback per step. Big-bang migrations fail in ways small-step migrations don't.
Full Prompt
An installable migration-plan skill. Reads the current code and changelog, then outputs ordered, reversible upgrade steps with a rollback note for every one.
What It Does
- Reads real usage — counts call sites, not docs
- Reads the upstream changelog — surfaces removed, renamed, and silently changed APIs
- Reversible steps — every step lands the repo in a working state
- Per-step rollback — one line each, not a promise of a runbook later
Install in 30 Seconds
Pick your tool above and download:
- Claude Code:
~/.claude/skills/migration-plan/SKILL.md - OpenAI Codex CLI: append to
AGENTS.md - Cursor: append to
.cursorrules
Same content, three install paths. Reload and ask for the audit on your next change.
Why This Skill Earns Its Keep
Big-bang migrations fail in ways small-step migrations don't. This skill enforces the small-step discipline: ship in clusters, keep the old and new in parallel where possible, leave the cleanup commit for last. If the upstream changelog doesn't exist, it tells you to wait — that's the right call.
Install once, run before every push.