Regression Check

---
name: regression-check
description: Use before moe.complete_task. Runs the suite the plan named — narrow for a mid-epic task, full for the epic's final task — to confirm nothing unrelated broke. The goal is zero regressions. Better to find out now than in a QA reject comment.
when_to_use: Worker, after the final implementation step is done, before moe.complete_task.
allowed-tools: Bash, Read
---

# Regression Check

Before `moe.complete_task` — **once, not after every step** — run the suite the plan named. The goal is zero regressions. Finding out now is cheap; finding out from a `qa_reject` is expensive.

## How wide, though?

The plan sizes this deliberately; don't silently upgrade it.

- **Mid-epic task** (there are sibling tasks after this one in the epic): run the narrow suite the plan named — your package, or the test files covering what you changed. You are proving *your slice*. The epic's final task owns the full-system pass.
- **Epic-final task, or a task with no epic siblings**: run the full thing. Every affected package, plus whatever integration/smoke the plan named.
- **Any task that touched shared types, schema, wire protocol, or a migration**: full suite regardless of position. These break code outside your blast radius, and the epic's final task is too late to find out.

Unsure of your position? `moe.list_tasks {epicId}` and compare `order`. No siblings after you → treat it as epic-final.

## What to run

The plan should name the suite. If it doesn't, work out from your changes:

| Touched | Run at minimum |
|---------|----------------|
| `packages/moe-daemon/src/...` | `cd packages/moe-daemon && npm test` |
| `packages/moe-proxy/src/...` | `cd packages/moe-proxy && npm test` |
| `packages/moe-claude-plugin/src/...` | `cd packages/moe-claude-plugin && npm test` |
| `moe-jetbrains/src/...` | `cd moe-jetbrains && ./gradlew test` |
| `moe-vscode/src/...` | `cd moe-vscode && npm test && npm run compile` |
| Multi-package or shared types | All of the above |
| Scripts / wrappers | `bash scripts/tests/postflight.sh` + `bash scripts/tests/parity-check.sh` |
| Docs only | `npm run lint` at the repo root (role-doc line caps) |

If a project has a `test:all` or `npm run check` script, prefer that — it usually wires lint + type-check + tests in the right order.

## How to read the output

- **All green?** Submit this run as `moe.complete_task { verification: { command, exitCode: 0, outputTail } }` (required) and put the test/pass counts in your summary. Don't claim green without numbers.
- **Failures in tests you didn't touch?** That's a regression. Investigate before `complete_task`. Usual suspects: shared util change, type-signature change, fixture / seed dependency, test ordering.
- **Failures in tests you did touch?** Either the test is wrong or the code is wrong. Fix one.
- **Flake?** Run it again. If it's still red on the second run, it's not flake, it's a bug.

## What "broader" means — when you're in the full-scope case

Once you're in the full-scope case above, it's tempting to still run only the tests in the package you edited. Don't. Cross-package dependencies are exactly where regressions hide. Specifically:

- Type changes in `schema.ts` ripple through every tool and the plugin.
- Migrations affect `load()` for every existing `.moe/`.
- Wire-protocol changes (new fields in WebSocket / MCP messages) need both ends tested.

## When to skip

- Doc-only changes (formatter / spell-check is enough).
- Pure addition of a new file that nothing imports yet.

For everything else, run *something* — narrow or full per the sizing above. Skipping entirely is what turns 90 saved seconds into a 4-hour `qa_reject` round trip.

## What to put in the complete_step / complete_task summary

Be specific. Not "tests pass" — "342 / 342 tests pass; ran `npm test` in moe-daemon and moe-proxy; type-check clean." Numbers + commands let QA verify quickly without re-running everything.