independent-review-loop

Operational spec for the team lead. This skill runs an independent review pass at delivery: a reviewer that fails differently from the author reads the whole PR against the approved outcome, the lead fixes what is in scope, and the loop repeats until the reviewer finds no more in-scope functional issues. It is the automated form of "ship, then loop the PR through Codex until it stops finding edge cases" — with the lead acting as the operator who keeps findings on-scope.

What distinguishes this from the team's own review and the recursive-refinement ladder is independence and exhaustiveness: the reviewer is a different model (Codex) or a fresh-context agent that fails differently from the authors, and it runs to exhaustion — until no in-scope functional finding remains — rather than the bounded rung ladder. (The ladder hunts bugs too, as it drives the work to the full scope of the outcome; the new axis here is the independent eye run to clean, not bug-hunting per se.) It is a distinct pass; it never replaces the ladder.

Only the lead runs this. Reviewers (Codex, or fresh subagents) are read-only; the lead is the sole writer, same as every other phase.

When this runs

Invoked from the unified pre-ship gate in code-mode's Refine/Deliver (and /swarm:refine) — the gate the lead presents once the team reaches 9/10+, offering (thoroughness-descending): recursive refinement + independent review / recursive refinement only / independent review loop only / ship as is. This skill runs for the two options that include the independent loop — recursive refinement + independent review (after the recursive ladder completes) and independent review loop only (on its own); recursive refinement only and ship as is do not invoke it. Deliver invokes this skill when the independent-review-loop pending run-state task recorded at the gate is open — it branches on that recorded task, not on recall of the pick.

It runs after the ship steps complete (PR-then-loop on the common path): when a PR was created, the loop reviews the PR's diff and pushes each round's fixes to it; on a commit-only / push-only ship with no PR, it reviews the pushed branch against its base. The diff base is resolvable from the PR when one exists (see the loop's base resolution below for the no-PR fallback).

Engine choice (Codex or Swarm fallback)

Persisted preference first. Before asking, read .claude/swarm-review.md — the per-project reviewer config. It mirrors the mechanism of .claude/swarm-ship.md (ask once, read thereafter) but is a separate file: reviewer config is not ship config. If it names a recognized engine (Codex or Swarm fallback), use it and skip both questions below — print a one-line plain-text notice so the choice is visible and overridable ("Using saved reviewer: — edit .claude/swarm-review.md to change"), no modal. If the file is present but its engine value is missing, garbled, or unrecognized, treat it as no saved preference and fall through to the questions below — a typo must never wedge the choice. A saved preference only selects the engine; it never asserts availability. Still run the runtime check (command -v codex + the first call's auth result). If the saved engine is unavailable on this machine, auto-degrade to the Swarm fallback without a modal — print a one-line notice ("Saved reviewer Codex unavailable here — using the Swarm fallback for this run") and do not rewrite the file. A saved preference must never become a blocking modal on a machine that lacks the tool — that is the portability guarantee (and it differs from an interactive Codex pick that's absent, which may ask). Swarm prompts are portable across machines, so a peer missing Codex neither hangs on a teammate's saved pref nor has it silently overwritten.

When no preference is saved, select the reviewer engine with AskUserQuestion (header "Reviewer"). Detect Codex once: command -v codex (exit 0 = present). Offer both engines every time — command -v codex only sets the order/default and a neutral note. No affordability language, ever. ("No loop" is not offered here — that is the gate's "ship as is.")

• Codex present → 1) "Codex" 2) "Swarm fallback".
• Codex absent → 1) "Swarm fallback" 2) "Codex (requires Codex CLI)".

Descriptions (neutral): Codex = "An independent model (Codex) reviews the whole PR; fails differently from the code's author." Swarm fallback = "A fresh Claude reviewer in a Codex style — sharp, but shares the author's model blind spots; always available, no Codex needed."

Then ask the scope of that choice (only when no preference was saved) with a second AskUserQuestion (header "Save reviewer"): "Use this reviewer just this session, or always for this project?" — "This session" (default — nothing is written) / "Always" (persist it). Only "Always" writes .claude/swarm-review.md, with a single ## Engine field naming the chosen engine:

# Reviewer Definition
## Engine
[Codex | Swarm fallback]

Persist the engine only — not availability, and not the rate-limit wait policy (the within-cap auto-wait below is default behavior, not a saved knob; the file is structured so a future field can be added without rework).

If the user picks Codex while it is absent/unauthed, surface Codex's own message + how to enable (codex login, or install per Codex's docs) once, then offer the Swarm fallback or stopping — never hang. While this gate is live and teammates exist, apply the launch.md live-team gate rules (ask teammates to hold; re-ask once if the modal is preempted).

The loop (identical control flow on both engines)

Resolve the diff base once. The review is simply the branch's committed diff against the branch it targets — main, or the repo's equivalent default/target branch. A GitHub PR is not required: Codex does a PR-style review of any branch, and the PR (when the ship definition uses one) is only the push target, not a precondition. Pick the base as the PR base if a PR exists (gh pr view --json baseRefName just names the target branch), otherwise the repo's default branch (git symbolic-ref refs/remotes/origin/HEAD --short, stripped of origin/); compare against the up-to-date base — the fetched remote-tracking ref (e.g. origin/main), not a possibly-stale local copy. The base must be the point the work diverged from, so the diff covers this run's changes; if the work was committed directly onto the base branch (so <base> and HEAD coincide), use the pre-work commit instead. If the base can't be resolved, ask the user to provide or confirm the base (or to explicitly opt out of the loop) — never silently skip the review the user selected, and never error. And if the resolved git diff <base>...HEAD comes back empty, that is NOT a clean review — nothing was reviewed — so surface it to the user to correct the base rather than terminating clean. Then each round:

1. Review the WHOLE PR diff, every round — the full git diff <base>...HEAD, never just the delta since the last fix. New findings hide behind old fixes; only a whole-PR pass surfaces them.

2. Collect findings in Codex's native shape (see Output format).

3. Relay the round to the user — verbatim, non-blocking, before the fixes. One contiguous plain-text block, one-way (emit-and-continue — never AskUserQuestion, never a wait for a reply; the loop does not stop), emitted once the triage in step 4 has determined the disposition and before any of the round's fix edits: a one-line header Round N — independent review, then the reviewer's output reproduced in full, verbatim — raw and unmediated, so the user sees the reviewer's voice before the driver's framing (on the Swarm fallback's multiple reviewers: each reviewer's output verbatim, labeled by its lens), then a one-line intent-phrased disposition — Fixing <n> in-scope; <m> filed out-of-scope, or Clean — no in-scope findings; loop terminating, or Reviewer returned no usable output — retrying. The disposition is forward-phrased ("Fixing…", never "Fixed…") — the fixes have not run when it renders. A round that ends in an escalation or degrade emits no leading block at all — the same-turn modal would collapse it; everything decision-relevant rides in the self-sufficient escalation modal (see Backstop and the Codex path), which quotes the reviewer's critical line(s) where a valid review ran and the verbatim error on a degrade. The disposition narrates the termination guard's determination (below); it never triggers or substitutes for it: write Clean … only when the guard actually cleared — the reviewer's own affirmative-clean line (relayed verbatim in this block; the lead never synthesizes "clean") or every finding triaged out-of-scope — and a FAILED review (empty/malformed/non-review) gets the retry line and no termination. This fires every non-escalation round a review actually ran, clean rounds included; an escalation round routes to the modal instead and holds the round's other in-scope fixes — every in-scope finding except the escalated one, where a specific finding is escalated — until the operator decision: the escalated finding follows that decision, never an auto-land, and on a continue the held fixes land, with a normal leading-block relay, on the resumed round, so the operator decides against a stable diff; a rate-limit park ran no review, so it emits nothing (on resume the round runs and the relay fires normally). It is distinct from the end-of-loop out-of-scope pile in step 5 — raw reviewer voice + this round's disposition here, the lead's cumulative triage ledger there; do not collapse them.

   IMPORTANT — the relay leads the round's rendering. Emit the relay as one contiguous leading text block ahead of any of the round's fix edits, in the same turn — never deferred to the turn's end, never split around the edits, never a back-reference ("relayed above"). Streamed text between tool calls renders live in a plain working turn; what collapses it is a same-turn AskUserQuestion — which is why the modal-ending escalation rounds carry their substance inside the modal itself (see Backstop and the Codex path) rather than in a leading block. As a thin durability net, recap each round's one-line disposition (never the verbatim reviewer output) in the final text block of a turn that ran review rounds. Run each round's review as a background invocation whose completion opens the next turn — one round per turn — so a later round's escalation modal can never share a turn with, and collapse, an earlier round's streamed relay; between-round boundaries are safe (the round's fixes are committed before the next review launches — nothing yields uncommitted state), unlike the rejected mid-round split between a round's relay and its fixes.

4. The lead triages each finding against the approved outcome — the lead is the codified "operator once in a while":

   • Bears on the outcome and is a real functional defect → fix it (lead writes the change) — except on an escalation round, where this round's fixes are held per step 3. Fix the class, not the site: a finding names one instance — before fixing, state the rule it violates and grep for sibling sites; fix the reachable siblings in the same round (a sweep that finds only the named site is a valid result, but it must be a result, not an assumption). When the same class recurs across rounds — new siblings, same rule — it is an invariant, not a bug; fork on this same triage, never pausing the loop: in-scope (the outcome requires the invariant everywhere) → stop scattering point-fixes, centralize the rule at one shared entry point, and surface the diff growth after per step 5's scope-growth discipline; exceeds the outcome → point-fix the named site now and file the enumerated remainder in the out-of-scope pile as a consequential item per step 5, recommending the chokepoint centralization as its resolution.
   • Out of scope (new feature, adjacent refactor, gold-plating, pre-existing issue, style) OR genuinely ambiguous → do not fix; add to the out-of-scope pile. When unsure, surface — never silently drop and never silently fix.

5. Surface the out-of-scope pile to the user (plain text) at the end of the loop, or immediately if a pile item is consequential (a real defect you're declining because it exceeds the approved outcome) — that is the rare operator escalation the user asked for. The diff may legitimately grow toward outcome-completeness (e.g., fixing an off-diff caller the outcome requires); that growth is bounded by the outcome, grounded in git diff --stat <base>...HEAD, and surfaced — not silent scope expansion.

6. Verify each fix landed, then commit this round's fixes (push per the ship definition), then re-run. Before committing, confirm each fix actually applied and changed the intended behavior — a scripted or multi-site replace that matched nothing is a silent no-op, not a fix, and a build exit code proves compilation, not behavior ("never claim compliance you didn't execute," applied to the lead's own edits). The review reads the committed git diff <base>...HEAD (the reviewer runs git locally), so the lead MUST commit each round's fixes (e.g., review: round N — <summary>) before re-running — otherwise the reviewer re-reads stale committed state, the loop never converges, and every just-fixed finding reappears unchanged and is misread as oscillation. Pushing is governed by the approved ship definition, not the loop: if the ship definition pushes (a PR or push workflow), push each round — to the PR if one exists, else the branch — to keep the remote reflecting the work; if it is commit-only, do NOT push (committing is enough for the review, and pushing would silently change the user's approved delivery or fail on a branch with no upstream). After committing (and pushing where the ship definition calls for it), re-run the review on the whole diff.

Termination (engine-neutral — identical on both paths) — stop when, after the lead's triage of a valid review, no in-scope functional findings remain. Two cases satisfy this and both terminate: (a) the review is an explicit, format-conformant affirmative "no findings" result; or (b) the review returned findings but the lead triaged them all as out-of-scope/ambiguous — terminate and surface the out-of-scope pile. Do NOT re-run a round whose only findings are out-of-scope (re-running an already-on-scope-clean state just churns to the backstop). What must NOT be read as termination is a non-review: empty output, a truncated or off-format response, a clarifying question, or a non-zero exit is a FAILED review, not a clean one — retry once, then degrade (Codex: take the error path / offer the Swarm fallback; Swarm: re-spawn the reviewer). Never read silence or a non-answer as "done": a false "done" silently ships bugs while reporting success, defeating the loop's whole purpose. (This is a validity check — was this a real, well-formed review? — combined with the lead's in/out-of-scope triage; not a field-by-field parser.) Both Codex and the swarm-native reviewers emit the same finding format, so the terminator never depends on an engine-specific field; on the Codex path an overall_correctness: patch is correct verdict, when present, is a convenience corroborator only, not the stop signal. Expect this to take many rounds — 8–10 rounds of genuinely different findings is normal and runs fully autonomously — but the stop condition is the clean review itself, not a round count: once a valid review returns no in-scope findings (case (a) or (b) above), terminate; do NOT re-run an already-clean review hoping to surface more. "Don't stop early" means don't declare done without a clean review (a small-but-nonempty round is not a clean review) — it never means re-running a review that already came back clean.

Backstop — a hard cap of 15 rounds (a fixed constant, not a config knob) guards against a non-converging loop. Reaching it does NOT silently abort: escalate to the user (continue / stop here / take over) — as do repeated invalid reviews (several consecutive failed/malformed reviews after retry/degrade), rather than silently riding the cap. An oscillating finding — the same finding, matched by its title + file:line (never by [P#], which is severity and is shared across unrelated findings), resurfacing after ≥2 fix attempts — is a disagreement, not a grind: stop auto-fixing it and surface it to the user with both positions for an operator decision. (A recurring class of new siblings is not oscillation — broaden the fix per step 4, never the oscillation key.) Escalation modals are self-sufficient: a same-turn AskUserQuestion collapses the turn's streamed text, so each escalation carries everything decision-relevant inside the modal's own content — and where a valid review ran, that content quotes the reviewer's own critical line(s) verbatim, never only the lead's paraphrase (the step-3 auditability, carried into the modal): oscillation quotes the resurfacing finding alongside the lead's position; the round-cap escalation includes what the reviewer is still finding, in its own words; and both name the held-fix count from step 3 — "also holding N in-scope fixes", N counting the held others and excluding the escalated finding — so a stop is chosen knowing what dies with it (a degrade modal names no held count: a failed review has no in-scope findings to hold). (Unlike a standing-team review dispute, this loop's reviewer is Codex or an ephemeral one-shot — not an addressable teammate — so swarm:resolve-dispute, which messages the reviewer, does not apply here; the user is the arbiter. This is the spirit of the "break review loops with evidence" rule, routed to the operator.)

Whenever the loop returns control to Deliver having terminated (it will not resume), Deliver closes the independent-review-loop pending task. That is EVERY terminal exit, not a select list: a clean review; all findings triaged out-of-scope; a rate-limit escalation resolved as stop; or the user choosing stop or take-over at ANY escalation — the 15-round backstop, an oscillation, repeated invalid/failed reviews, a Codex auth/fatal degrade the user ends, or an unresolvable/empty base the user opts out of. What does NOT close it is a non-terminal pause that will resume: a timed park, an escalation still awaiting the user's decision, or a fallback / wait-anyway / continue. An open task after a terminal return would re-churn the loop via the pulse; closing it during a pause would lose the loop. (The test is mechanical: will the loop run again? Pause → keep open. Done → close.) The terminal handshake never shares a turn with a round relay: when the loop terminates, the turn ends with the final round's relay, the out-of-scope pile, and every in-session delivery artifact the ship type owes (the no-PR Decision digest; /swarm:refine's in-session summary) as its final text block — Deliver's in-session content completes here, so only the terminal handshake (pulse-delete, then the Terminal gate) runs on the next wake, which the pulse's own terminal backstop already implements for both ship classes once the in-session artifact has landed; a same-turn Terminal modal would collapse the loop's final relay, the exact loss this relay exists to prevent. (A terminal-turn boundary is safe where a mid-round one is not: the loop will not resume, so nothing races the pulse.)

Auditability — why step 3 relays raw, before the fixes: the lead is both loop-driver and fixer, so the user — not the self-interested driver — must be the backstop on "are we actually done," and a false Clean disposition is visibly contradicted by the reviewer's own text sitting directly above it. The obligation itself lives in step 3; this note is rationale only.

The steer (one string, both engines — this is the over-reach guard)

Send the reviewer exactly this (substituting the approved outcome verbatim and the resolved base):

The outcome this change is meant to achieve: «approved outcome, verbatim»

Review the ENTIRE change on this branch relative to «base» — the full PR diff (git diff «base»...HEAD), not just the latest edits. Report only material, FUNCTIONAL findings that bear on that outcome: incorrect logic, defects, broken or unhandled edge cases, and regressions in code this change touches.

Do NOT report: style, naming, or formatting; low-value cleanup; pre-existing issues this change did not introduce; speculation you cannot tie to a concrete failing input or condition; or choices the author clearly made intentionally. A finding must be discrete, actionable, and something the author would fix if they knew.

Prioritize each finding [P0]–[P3]. No code suggestions — for each finding, explain in prose what breaks, the exact conditions that trigger it, and the consequence. Format: open with a 1–3 sentence summary, then a line Full review comments:, then one bullet per finding as - [P#] <imperative title> — <file>:<line-range> followed by the indented prose paragraph. If you find no in-scope functional issues, state that explicitly (e.g., "No in-scope functional findings.") — do not return an empty or merely terse response.

The closing affirmative-clean instruction is load-bearing: it is what lets the termination guard above distinguish a genuinely clean review from a malformed/empty/aborted one. Without it a clean reviewer may return nothing, the guard reads that absence as a FAILED review, and the loop never terminates — burning to the round cap and firing a spurious "didn't converge" escalation on work that was actually clean.

Telling the reviewer the outcome and asking only for functional findings is the steer that keeps the loop on-scope — the input-side guard, paired with the lead's triage on the output side.

Codex path

Codex is the preferred reviewer because it is a genuinely different model and fails differently from the code's author. Run it via Bash (the lead can run shell; /codex:* slash commands are not model-callable, so do not rely on them):

• Invoke with the steer as a custom review prompt passed on STDIN — never as a double-quoted shell argument. The steer contains backticks (e.g. `git diff …`, `codex login`); inside double quotes bash runs them as command substitution and corrupts the prompt. Both codex review and codex exec review accept the prompt from stdin via the - positional, so write the steer to a temp file (or a single-quoted heredoc) and pipe it: printf '%s' "$steer" | codex exec review -. (Codex's review flags --base/--uncommitted/--commit are mutually exclusive with a custom prompt, so the steer rides as the stdin prompt and itself names the git diff «base»...HEAD scope; codex still applies its built-in review rubric as the system prompt.)
• Capture stdout, stderr, and the exit code explicitly — codex exec exits non-zero on a fatal error, which is the degradation path below. If the invocation errors, check codex exec review --help / codex review --help and adapt — do not hardcode against version drift. The pass is read-only; if codex's default sandbox blocks the git the steer relies on, surface that and degrade rather than hang (runtime-validate).
• Depend only on the codex binary, never on the codex plugin, its install path, or its companion script.
• Auth/fatal errors degrade, never hang: the first failing call that is not a usage-limit error (auth, sandbox, or other fatal — usage limits are the separate auto-wait class below) → surface Codex's verbatim error + codex login, then offer the Swarm fallback or stopping (the verbatim error rides inside the offer modal itself — the self-sufficient-modal rule in Backstop; a failed review has no reviewer line to quote). Exception — a saved-preference engine: if Codex came from a saved .claude/swarm-review.md preference (not an interactive pick this run), auto-degrade to the Swarm fallback without a modal (one-line notice, no file rewrite) — the same portability guarantee as a missing CLI, covering the present-but-unauthed case too; only an interactive Codex pick gets the offer-fallback-or-stop modal.
• Usage-limit (rate-limit) errors auto-wait — a distinct class, not a generic failure. A Codex usage-limit error is transient and self-healing: it carries a reset time (resets_at, or "try again at " / "try again in "). Codex has two windows — a rolling ~5-hour window and a weekly one — and the CLI does not auto-resume, so the loop must. The whole wait state lives in ONE durable run-state task — never in volatile context — so the attempt count survives a compaction and the pulse (which resumes the round post-compaction) can read and enforce it. Distinguish a usage-limit error from auth/fatal/stall by its message/class, then:
  1. Park on a durable task. Commit the current round's fixes first (never yield a dirty working tree — there must be a clean state to resume onto). On the FIRST park, create one codex-reset-wait task whose description carries until <ISO reset+small buffer>, since <T0> (the first-park time, set once and never overwritten), attempt 1, state parked-on-timer, and the rules the pulse follows: hold until until, then resume the SAME loop round in place, plus the attempt/escalate bound below. Announce once in plain text ("Codex usage limit; resets ~HH:MM, ~ min — waiting, will auto-resume; reply to switch to the Swarm fallback now if you'd rather not wait"), then yield. The existing 20-min pulse resumes the round after the reset (≤20-min latency; no second cron). Do not block the session foreground — foreground sleep is blocked, and a blocking wait would freeze the team and be preempted by the pulse anyway.
  2. Re-park updates the SAME task — never close-and-recreate. If Codex is still limited on resume, TaskUpdate the existing task: new until <T>, increment attempt <N>; leave since <T0> untouched. The one task is the entire park sequence's durable record — close-and-recreate would reset the counter and reopen the infinite-park bug.
  3. Bound the single wait AND the sequence — and route EVERY escalation through the task. All three escalation triggers — an immediate beyond-cap reset (a weekly window is days away), an unparseable reset, and (from the durable state) 2 non-clearing attempts or cumulative wait (now − T0) beyond the cap — resolve the same way: set the codex-reset-wait task to state escalated-awaiting-user (creating the task first if this is an immediate beyond-cap/unparseable escalation that never parked), surface the choice once, and hold. This gives every escalation the same durable, compaction-surviving "surface once then hold" as a timed park — the pulse reads the state and holds for the decision rather than re-asking per cycle or self-deleting. Escalation presents the user wait-anyway / Swarm fallback / stop — it never silently swaps the user's chosen/persisted engine (auto-substitution would override the engine preference the user set; out of scope). If the user is away at a genuine cap-exceeded, the loop legitimately stalls here.
  4. Close only on real resolution. Close the task (TaskUpdate → completed) ONLY when Codex actually clears (a real review ran that round), an escalation resolves (fallback or stop), or the user switches to the Swarm fallback within-cap. NEVER close on a bare "resume" — a resume that re-hits the limit keeps the same task and increments the attempt; closing on resume is exactly the non-durable bug that resets the counter.
  5. Resume in place — don't restart. An open codex-reset-wait means this loop round is parked, not finished: continue the SAME round against the already-resolved base — do not re-enter the loop from the top, do not re-resolve the base, do not reset the 15-round backstop. While it is open, the independent-review-loop pending task is satisfied by this in-progress (parked) loop — it is not a signal to start a fresh run.
  6. Resolving an escalation (and the within-cap "reply to switch") — reuse close-every-exit, don't invent new close logic. When the user (or a within-cap reply) resolves the wait, route by the choice:
     • fallback (the escalation pick, or a within-cap reply with clear switch-intent) → close the codex-reset-wait task, switch the reviewer engine session-scoped only (do NOT rewrite .claude/swarm-review.md), and resume the SAME round on the fallback. One path for every way the engine switches.
     • wait-anyway → re-park with a FULL fresh window: new until <T>, new since <T0>, attempt 1, state parked-on-timer, staying on Codex. This is the one authorized exception to the set-once since contract (the run-state vocabulary names it) — an explicit user wait-anyway starts a new sequence, which is why since <T0> is reset here. until = the parsed reset when one is known; for an unparseable reset (none was ever parsed), default until to now + one rolling window (~5h) so the wait stays bounded rather than open-ended. (A partial reset that kept since <T0> would let the cumulative-wait bound re-escalate almost immediately, making "wait anyway" hollow for an away user.)
     • stop → close BOTH codex-reset-wait and independent-review-loop pending, terminate the loop, and surface to the user that the independent review stopped incomplete. Within-cap switch detection is conservative: switch only on an unambiguous switch-intent reply; an ambiguous or unrelated reply passes through untouched (no silent switch), and a switch gets a one-line ack so a present user knows it landed. These are the rate-limit additions to the loop's close-on-every-terminal-path discipline (see the Backstop section) — not separate close machinery.
• Codex's own Stop review-gate (/codex:setup --enable-review-gate) is a separate, unbounded per-turn loop that will collide with this one. Warn the user once in prose to disable it for this run; do not read or modify the user's Codex config.
• Review depth tracks the user's Codex config, don't hardcode it. The 8–10-round expectation assumes a strong reviewer (the practice this automates runs the best available model at high/xhigh effort); let the user's codex model/effort settings drive depth rather than pinning a model in the invocation.
• Guard against a stalled invocation, not just a failed one. A review that hangs (no progress, no exit) is distinct from a non-zero exit — bound each call with a no-progress/timeout cutoff and treat a stall as a failed review (degrade per above), so one hung review can't hang the loop. (Applies to any review invocation, including a swarm-fallback reviewer that goes dark.)

Swarm fallback

When Codex is absent — or the user picks it deliberately — run the same loop with fresh, ephemeral, read-only reviewer subagents spawned via the Agent tool (subagent_type: swarm-reviewer, the plugin's ephemeral reviewer definition; pass no name, so each reviewer runs as a plain sub-agent whose findings return to the lead when it completes), never as team members joining the standing team. Fresh per round: each reviewer's spawn prompt MUST carry the full steer verbatim — the same string the Codex path uses, including the affirmative-clean instruction and the output format — plus the diff and the approved outcome, with no Converge/Approve history. This approximates an outside reviewer and avoids the "I already looked, looks fine" fatigue of re-soliciting standing members. Verify the affirmative-clean line is actually in the spawn prompt (don't assemble the spawn from persona + a loose paraphrase that drops it): it is what lets the termination guard tell a genuinely clean swarm review from a malformed/empty one; if it's dropped, a clean fallback review never terminates — the exact regression this guards against. Spawn serially (one reviewer at a time); use distinct lenses across reviewers (e.g. correctness & logic, edge cases & failure modes, regressions in touched code). They return findings only; the lead triages and fixes exactly as on the Codex path.

Each reviewer's identity is a Codex-style reviewer (adapt this as the subagent's identity, not as extra brief sections):

a code reviewer in the style of Codex — terse, analytical, matter-of-fact, zero flattery or praise. Flags only discrete, actionable, functional defects introduced by the change that the author would fix if they knew; states the exact conditions that trigger each bug and the provably-affected code; never speculates and never nitpicks style. Reviews the whole PR against the stated outcome and reports findings in the prescribed format with no code suggestions.

Honest limit, worth stating to the user: a fresh Codex-style Claude reviewer is sharp and adversarial but shares the model-level blind spots of the team that wrote the code — it buys stance, not the independence Codex's different model gives. That is why Codex is offered first when present.

Output format (both engines — Codex's native shape)

Findings are rendered as:

«1–3 sentence overall summary of the themes»

Full review comments:

- [P1] «imperative finding title» — path/to/file.ext:120-128
  «one prose paragraph: what breaks, the exact conditions that trigger it, the consequence, and what to do — no code blocks»

- [P2] «next finding title» — path/to/other.ext:40-44
  «…»

[P#] is a presentation/severity label only. It is neither the action gate (what the lead fixes is decided solely by the in-scope/out-of-scope triage above, anchored on the outcome) nor the oscillation key (recurrence is matched by finding identity — title + file:line).

What this skill reuses (do not rebuild)

The round structure, batch-fix-then-re-review discipline ("wait for all of a round's findings before fixing"), escalation, and the serial cadence are all existing swarm machinery — this skill only adds the independent reviewer, the outcome steer, the per-round verbatim relay, the triage gate, the termination/backstop, and the output contract. (Oscillation routes to the user, not swarm:resolve-dispute — that skill messages a teammate reviewer, and this loop's reviewer is Codex or an ephemeral one-shot.) No new phase, no new hard rules, no changes to the team-member briefing templates.

---

The reviewer rubric, priority tags ([P0]–[P3]), finding format, and steer are adapted from OpenAI Codex (codex-rs, Apache-2.0), with one deliberate deviation: Codex permits short suggestion blocks; swarm requires prose-only findings with no code suggestions.