pan-review git:20260722.6605404A

Immutable. This exact content is served forever at /api/v1/blob/ef3029a53df888d8.

---
name: pan-review
description: "pan review <subcommand> — manage the code review lifecycle: list pending work, re-request review, heal status drift, reset/abort/restart review cycles"
triggers:
  - pan review
  - review pending
  - request review
  - review mode
  - reset review
  - resync review
  - restart review
  - abort review
  - code review lifecycle
allowed-tools:
  - Bash
  - Read
---

# pan review

Manage the review pipeline for completed agent work. Use this when an issue
has been signaled done by the work agent but you need to inspect, retry, or
abandon the review/test/ship pass.

## Usage

```
pan review pending                                 # List completed work awaiting review
pan review pending --ready                         # List issues ready for merge
pan review pending --blocked                       # List blocked review/test/merge issues
pan review request <id>                            # Re-request review after fixing feedback
pan review mode <id> <quick|full>                  # Set this issue's review mode override
pan review reset <id> [--session]                  # Reset review/test/merge cycles (human override)
pan review resync <id>                             # Re-emit canonical status to heal dashboard drift
pan review abort <id>                              # Kill all running reviewers, leave worker idle
pan review restart <id> [--model <m>] [--role <r>] # Kill reviewers and dispatch a fresh review pipeline
```

## What each subcommand does

- **`pending`** — Uses durable pipeline membership as the issue universe, then
  shows review state as an annotation. The default view lists pending reviews
  plus `zombie_pr` and `post_merge_limbo` drift; with no pending review rows it
  still prints the resolver-derived pipeline set. `--ready` and `--blocked`
  preserve their review/test/merge meanings, filtered to pipeline members.
- **`request <id>`** — After fixing the issues a reviewer flagged, this
  re-triggers the review pipeline against the current branch state. Use this
  when the worker has committed fixes and you want the existing review pass
  to re-evaluate.
- **`mode <id> <quick|full>`** — Sets a per-issue review-mode override in the
  issue record. `quick` runs the single parent review agent. `full` opts the
  issue into the convoy review: security, correctness, performance, and
  requirements sub-reviewers plus synthesis. Use this only when heavier review
  is worth the extra agent/runtime cost for this one issue.
- **`reset <id>`** — Clears the review/test/merge state for an issue, so the
  pipeline can be re-dispatched from scratch. Use when the saved state is
  inconsistent or corrupt. `--session` additionally clears the saved Claude
  session for each reviewer so they restart with a clean conversation.
- **`resync <id>`** — Reads the canonical review status and re-emits it without
  changing any verdict. Use when an issue passed review but is missing from
  Awaiting Merge or its actions are wrongly gated by stale dashboard state.
- **`abort <id>`** — Kills any currently running reviewer sessions but
  leaves the work agent alone. Use when reviewers are stuck or running
  against the wrong commit and you want to halt without resetting state.
- **`restart <id>`** — `abort` + dispatch a fresh review pipeline in one
  command. Use when reviewers crashed or produced unusable output and you
  want a clean re-run. Optional flags:
  - `--model <model>` — override the model for every reviewer in this run
    (e.g. `gpt-5.4`, `claude-sonnet-4-6`). Useful when the default model has
    misbehaved and you want to retry with a different one.
  - `--role <role>` — restart only one reviewer role (`correctness`,
    `security`, `performance`, `requirements`) instead of the whole convoy.

## When to use each

| Situation | Command |
|---|---|
| "What's waiting on me?" | `pan review pending` |
| "What's ready to merge?" | `pan review pending --ready` |
| "What's blocked in review/test/merge?" | `pan review pending --blocked` |
| Worker pushed a fix, want re-review | `pan review request <id>` |
| This issue needs the full review convoy | `pan review mode <id> full` |
| Return an issue to default quick review | `pan review mode <id> quick` |
| Pipeline state is inconsistent, need a clean slate | `pan review reset <id>` |
| Issue passed but is missing from Awaiting Merge or actions are wrongly gated | `pan review resync <id>` |
| Same as above plus reviewer Claude sessions are bad | `pan review reset <id> --session` |
| Reviewer is hung, just kill it | `pan review abort <id>` |
| Reviewer crashed, want a fresh convoy with a different model | `pan review restart <id> --model gpt-5.4` |
| Only the security reviewer is broken | `pan review restart <id> --role security` |

## Merging is NOT here

`pan approve` has been removed. To merge an approved branch, use the
**MERGE** button on the dashboard. The merge agent runs autonomously after
all reviewers + the test specialist sign off; humans only click the button.

## See also

- `pan show <id>` — inspect work agent state, xBRIEF status, recent activity
- `pan done <id>` — signal initial work completion (from the worker side)
- `pan code-review` skill — orchestrated parallel code review with synthesis
- `roles/review.md` — the review role's frontmatter and prompt
- `docs/REVIEW-AGENT-ARCHITECTURE.md` — full design of the review convoy