review-loop git:20261003.6785f65A

Immutable. This exact content never changes and is served at /api/v1/blob/11a1fd36c13ea569.

---
name: review-loop
argument-hint: "<work item description> [--handsfree]"
description: >
  Automates a Plan-Execute-Review workflow with built-in iteration loops.
  An Orchestrator coordinates an Executor sub-agent and a configurable Reviewer
  (external AI CLI like Codex, or Claude sub-agent) to drive work items from
  description to delivery — with every review finding visible to the user.
  Trigger: "run review-loop on", "start review loop", "let the agents handle",
  or any task the user wants driven by a plan→review→implement→CR cycle.
---

# review-loop — claude orchestration

Read `docs/protocol/loading.md`, then run:

```bash
python3 <support-root>/scripts/read_protocol.py --runtime claude --stage entry-review-loop --output .review-loop/tmp/protocol-entry-review-loop.md
```

Resolve <support-root> to this plugin/repository, not the task workspace.
Keep cwd in the user's workspace. After exit 0, read the complete output file
in bounded chunks before acting; stdout contains only a compact hash/size receipt,
not the instruction body. A missing, unreadable, or incompletely read file blocks the action.
Run every loader call as its own Bash command, never chained (rule in `docs/protocol/loading.md`).
`docs/protocol/loading.json` is the action/prerequisite map for both runtimes.

Fresh work may hand off to paired-session (the default when `entry` is absent) per the entry procedures.
Detect fresh / plan-exists / code-exists / explicit-resume via the entry procedures. For fresh work run planning → planning-review; after approval continue execution → execution-review → gate → polish → docs → security → delivery. Do not stop after exec alone.

Before initialization load `session-init` AND `review-loop-init`; before a resume
load `resume` AND `review-loop-init`. Before any work-agent dispatch load the
matching `planning` or `execution` bundle; before review load
`planning-review` or `execution-review`. On planning approval, only the plan-only entry loads `plan-exit`.
The umbrella retains its session/lock and continues directly into execution.
Before a stage transition, retry,
stop or error exit, load its applicable bundle per `loading.md`.
Use `parallel-review` only in the Codex orchestrator for N>1 Claude-CLI
reviewer jobs (never in Claude Code or for local Codex Reviewer agents); load
`context-persist` only when that optional substep is applicable.

The caller owns the session file and lock. Preserve unrelated dirty work;
read-only/plan-only scope and user authorization override implementation steps.
Independent Reviewer approval, output validation, rubric/triage and evidence
guards remain mandatory. Use existing configured backend/model resolution.
Never infer a PASS, a skip, or permission from not loading a future stage.

Full session schema: `docs/protocol/session-file.md`; active loops:
`docs/protocol/planning.md` and `docs/protocol/execution.md`; output contracts:
`docs/protocol/executor-output.md` and `docs/protocol/reviewer-output.md`.
These are scoped references, not an eager import list. Detailed native entry
steps live in `references/entry.md` and are selected by the loading map.

Reuse a rule unit only while its exact text remains in this live context,
using the fingerprint emitted by the loader. Fresh agents, compaction and
new invocations must load their own prerequisites. Do not use persisted session
state as proof that instructions are still available.