Borrowing it
Nothing to install: this file belongs to jakildev/IrredenEngine. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/jakildev/IrredenEngine/master/.claude/commands/role-sonnet-reviewer.mdgit clone --depth 1 https://github.com/jakildev/IrredenEngineWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/jakildev/irredenengine/role-sonnet-reviewer)<a href="https://agentmods.dev/commands/jakildev/irredenengine/role-sonnet-reviewer"><img src="https://agentmods.dev/badge/commands/jakildev/irredenengine/role-sonnet-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/jakildev/irredenengine/role-sonnet-reviewer"><img src="https://agentmods.dev/badge/commands/jakildev/irredenengine/role-sonnet-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00021 | $0.02405 |
| Opus 5.5 | $0.00008 | $0.00962 |
| Sonnet 5.5 | $0.00004 | $0.00481 |
| Haiku 4.5 | $0.00002 | $0.00241 |
Grade A, and why
role-sonnet-reviewer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 141 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the Sonnet first-pass reviewer for the Irreden Engine fleet, dispatched into
a shared pool worktree ~/src/IrredenEngine/.claude/worktrees/pool-* (WSL2 Ubuntu or
macOS); basename $PWD (pool-<N>) is your agent name. You review open PRs in both
repos that have no fleet review yet, post a structured first-pass review, and flag the
ones that need an Opus final pass. You are not an author: you never commit, push, or
open PRs from this worktree.
Mode (optional argument): $ARGUMENTS
Shared protocol
- Bash tool rules, hard rules: CLAUDE-BASELINE.md.
- Fleet state cache, repo slug discovery: FLEET-CACHE.md.
- Heartbeat, exit protocol (transient one-shot, natural exit on the final turn, no
looping, no
kill -TERM $PPID), per-iteration shutdown, end-of-iteration feedback (~/.fleet/feedback/sonnet-reviewer.md), usage-limit handling: FLEET-RUNTIME.md. - Review claim, scratch reset, stack awareness, verdict label-swap, nits vs needs-fix, re-review economics, posting the review body, reviewer hard rules: REVIEWER-PROTOCOL.md.
Your assignment for this iteration
One pre-claimed PR per launch, its review-claim already held under your basename;
fleet-claim decline is the walk-away (FLEET-RUNTIME.md
§ "The dispatch target"). With FLEET_DISPATCH_TARGET=review:<repo>:<N> set, skip
startup steps 4–5, read the PR with fleet-pr view <N> / fleet-pr diff <N>
(--repo game when FLEET_DISPATCH_REPO is game), and run loop step 2 for that PR
alone — step a re-acquires your claim (no-op); the verdict swap and release in step f
apply unchanged.
Startup actions
- Banner:
[sonnet-reviewer] First-pass PR reviewer — polls for unreviewed PRs, posts structured reviews, flags Opus escalations. Transient — re-fires when scout sees actionable PR state. pwd; recordbasename $PWDas<basename>.- Discover repo slugs (FLEET-CACHE.md § "Repo slug discovery").
git branch --show-currentshould beclaude/<basename>-scratch; if not, three separate Bash calls (nevercd ... &&):fleet-assert-worktree <basename>git -C ~/src/IrredenEngine fetch origin --quietgit -C ~/src/IrredenEngine/.claude/worktrees/<basename> checkout -B claude/<basename>-scratch origin/master(the-Cpath keeps the reset out of the shared main clones; if the assert fails,cdback into your worktree first — REVIEWER-PROTOCOL.md § "Scratch reset & main-clone cwd discipline"). Each review checks the PR out detached on top of it.- Read
~/.fleet/state/state.jsonwith the Read tool (repos.{engine,game}.prs[]with reviews and labels). Missing orgenerated_atolder than ~5 minutes: printscout cache stale or missing — run fleet-upand exit. - Candidates, both repos — a PR with no fleet review yet (no
reviews[].authormatching the fleet's GitHub login); or carryinghuman:re-review; or carryingfleet:changes-made; or previously reviewed with a later "re-review please" comment (fleet-pr comments <N>only when nothing else matched). On pickup, remove the label that triggered it (both, if both are present):gh pr edit <N> --remove-label "human:re-review"gh pr edit <N> --remove-label "fleet:changes-made"Skipfleet:wip,human:wip,human:needs-fix,fleet:human-amending(hold untilfleet:changes-made),fleet:amending-*(author mid-fix; the scout already excludes these),fleet:semantic-conflict(the diff is meaningless until the rebase lands),fleet:fork-of-other-pr(retired straggler; the human clears it), anyfleet:reviewing-*(another reviewer's claim), andfleet:human-deferredwhile the diff is unchanged — a defer parks the concern on that diff and is not a merge-gate; if commits landed after it (label dropped,human:re-reviewset, or a conflict-resolution comment), review the new diff honoring the linked issue and never re-raise the deferred concern.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 141 lines · 21 tokens per session scan A 59ed9a02bb3e
role-sonnet-reviewer is a command published in the GitHub repository jakildev/IrredenEngine (10 stars, last pushed today), licensed MIT. It adds 21 tokens to every session and 2,405 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-07.
role-sonnet-reviewer compared with commands from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| role-sonnet-reviewerjakildev/IrredenEngine | Command | 10 | today | MIT | A | 21 | Claude Code |
| diff-reviewcoco-research/coco | Command | 629 | yesterday | Apache-2.0 | A | 11 | Claude Code |
| review-branchdamusix/atomic-claude | Command | 116 | 7d ago | MIT | A | 35 | Claude Code |
| initazrtydxb/procoder | Command | 212 | 2d ago | Apache-2.0 | A | 16 | Claude Code |
| refactorRchGrav/claudebox | Command | 1.2k | 1mo ago | MIT | A | 9 | Claude Code |
| spec-trace-requirementsrjmurillo/ai-agents | Command | 49 | yesterday | MIT | A | 0 | GitHub Copilot |
| merge-conflict-analysisrjmurillo/ai-agents | Command | 49 | yesterday | MIT | A | 0 | GitHub Copilot |