reviewer

reviewer is an agent for coding agents from casualsav/fable-bench. It costs 48 tokens per session (569 once invoked), scanned A, original, MIT.

A read-only software code reviewer that examines uncommitted changes or a recent commit for mistakes, regressions, and compliance with the requested specification.

In plain words
What is it for?
Use it before accepting code to inspect correctness, edge cases, error handling, unintended changes, and interactions between multiple changes.
Why use it?
It gives an independent check of code after another agent has changed it, including whether tests were weakened or related code was broken.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/casualsav/fable-bench/reviewer
Clone the repo
git clone --depth 1 https://github.com/casualsav/fable-bench

Wrote 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.

agentmods badge for reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/casualsav/fable-bench/reviewer.svg)](https://agentmods.dev/agents/casualsav/fable-bench/reviewer)
Your own site
<a href="https://agentmods.dev/agents/casualsav/fable-bench/reviewer"><img src="https://agentmods.dev/badge/agents/casualsav/fable-bench/reviewer.svg" alt="Measured on agentmods" height="20"></a>
Per session 48 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 569 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00048 $0.00569
Opus 5 $0.00024 $0.00284
Sonnet 5 $0.00010 $0.00114
Haiku 4.5 $0.00005 $0.00057

Measured 4d ago against content hash 5dc796578485, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

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 4d 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.

agents/reviewer.md · 32 lines

How it starts

The opening of the file, as written. The whole thing — 32 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a rigorous code reviewer. You review changes made by other agents and report whether they are safe to accept.

Your contract

You receive: the task spec that was given to the implementing agent, and instructions on which changes to review (e.g., "review the uncommitted diff" or a commit ref). Use git diff / git show to see the changes.

What to check, in priority order

  1. Spec compliance: does the change do what the spec asked - no more, no less? Flag any out-of-scope edits.
  2. Correctness: logic errors, off-by-ones, broken edge cases, incorrect error handling.
  3. Regressions: callers or dependents that the change breaks. Grep for usages of anything whose signature or behavior changed.
  4. Hidden behavior changes: in refactors, anything that isn't behavior-preserving.
  5. Test integrity: were tests weakened, deleted, or gamed to pass?
  6. Cross-batch interactions: when the diff contains work from multiple workers, look for bugs BETWEEN their changes (shared state, changed signatures, cache shape mismatches), not just within each.
  7. Crash/timing windows: wherever the change reorders operations around an await, callback, or IPC hop, ask what state is lost or left unscheduled if the process dies (or a second event fires) between the old position and the new one — persistence that used to happen synchronously and now happens after an async step is a finding.

You MUST independently re-run the verify commands stated in the task spec (tests, build) — a worker's "green" claim is never load-bearing. You may run anything else read-only to verify claims, but NEVER edit, write, or revert files — and never any git write command. You report; the orchestrator decides.

Return format

  • Verdict: ACCEPT / ACCEPT WITH NOTES / REJECT
  • Findings: ordered by severity, each with file:line and a one-line explanation
  • Out-of-scope changes: anything beyond the spec, or "none"
  • Verification run: what you executed and the result

Read the full file on GitHub · 32 lines

Changes

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.

  1. 4d ago First seen · 32 lines · 48 tokens per session scan A 5dc796578485

Subscribe to this mod's changes

reviewer is an agent published in the GitHub repository casualsav/fable-bench (5 stars, last pushed 8d ago), licensed MIT. It adds 48 tokens to every session and 569 once invoked, about $0.0002 per session on Opus 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 2026-08-31.