| name | self-review |
| description | Final check on your own changes before reporting a task done, opening a PR, or asking for review. Spawns a subagent that runs `/code-review` on the current branch — clean context, no "I just wrote this, it must be good" bias. Returns SHIP / ITERATE verdict. Triggers include "self review", "review my changes", "check my diff", "ready to commit", "ready for PR", "I'm done", "before I ship". |
| allowed-tools | Agent |
| metadata | {"version":"4.0","domain":"development","complexity":"intermediate","tags":"review, quality, pre-pr"} |
Self-Review
Step 1 — spawn the reviewer
Run this exact Agent call. run_in_background: true is non-negotiable — foreground blocks the harness.
Agent({
subagent_type: "general-purpose",
description: "Self-review current branch",
run_in_background: true,
prompt: "Run /code-review on the current branch. Return the full review verbatim plus a one-line SHIP or ITERATE verdict at the end."
})
While it runs, do not poll — you will be notified on completion.
Step 2 — report back to the user
When the subagent returns, your reply MUST have two sections:
What the reviewer said — paste the subagent's findings verbatim (or a faithful summary if very long). Do not silently drop items you disagree with.
My recommendations — for each finding, state: address / defer / ignore, with a one-line reason. End with your own SHIP / ITERATE call.
The user decides what to act on. Your job is to surface everything the reviewer raised, then advise.