← ClaudeAtlas

eng-reviewlisted

Engineering review of a plan or change on six dimensions 0-10 with what a 10 looks like, then repair.
rohithkandula19/Ronin · ★ 1 · AI & Automation · score 73
Install: claude install-skill rohithkandula19/Ronin
# eng-review — the engineering pass Pairs with `design-review`: that one judges whether the plan is the right shape, this one judges whether it is built like production. Run it on a concrete plan or a diff, from the `plan` role (switch with `/model`). Read-only — `read`, `grep`, and `glob` to inspect, and `bash` for non-mutating facts (`git diff`, `git log`, a dry-run build). The output is a scored critique and the fixes it implies, not edits. Score each dimension **0-10**: give the score, the one thing capping it, and **what a 10 looks like** in concrete terms for this change. 1. **Correctness & edge cases** — right answer on the boundary inputs, not just the happy path. *10: off-by-ones, empty inputs, and concurrency are each handled or explicitly ruled out.* 2. **Error handling** — what happens when a dependency fails. *10: every failure has a path that surfaces a usable message and leaves state consistent; nothing is swallowed.* 3. **Interfaces & naming** — the seam other code will bind to. *10: names say what they mean, the contract is minimal and hard to misuse, and callers cannot reach past it.* 4. **Performance & resources** — cost in time, memory, I/O, and handles. *10: no accidental N+1 or unbounded growth; the hot path is known and acceptable; resources are released.* 5. **Testability & coverage** — is the behaviour actually pinned? *10: the change ships with tests that fail before it and pass after, at the level where regressions occur.