agent-code-reviewer

An automated code-review agent that checks a pull request against its requested specification and then reviews its correctness, tests, and maintainability. A pull request is a proposed set of code changes for review before it is merged.

In plain words
What is it for?
It is for reviewing pull requests after implementation or before creating one. Security review is handled by a separate tool.
Why use it?
It separates checking whether the requested work was done from checking whether the code is in good shape. It returns a pass-or-reject decision and does not change the code.

Agent

Part of the maverick plugin — 41 skills, 5 agents, 5 hooks shipped together

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/thermiteau/maverick/agent-code-reviewer
Clone the repo
git clone --depth 1 https://github.com/thermiteau/maverick

Or install maverick, the plugin that ships this one along with the rest of its 41 skills, 5 agents, 5 hooks.

Per session 66 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,948 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 $0.00066 $0.01948
Opus 5 $0.00033 $0.00974
Sonnet 5 $0.00013 $0.00390
Haiku 4.5 $0.00007 $0.00195

Measured 3d ago against content hash c18e3453cfb4, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-code-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 3d 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/agent-code-reviewer.md · 206 lines

How it starts

The opening of the file, as written. The whole thing — 206 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Senior Code Reviewer. Your output is a binary verdict that decides whether a pull request is merged or ejected for human handling.

Contract

Input: a pull request URL (e.g. https://github.com/owner/repo/pull/173). The caller resolves the spec (issue body + design + task list) and passes those to you alongside the URL.

Two-stage review:

  1. Spec compliance — does the PR implement what the GitHub issue asked for?
  2. Code quality — is the implementation correct, well-tested, and maintainable?

Output: one of exactly two verdicts, terminated by the machine-parseable marker line (see Output format).

Read-only: you are a gate, not a fixer. You have no file-editing tools, and your gh/git usage must be read-only — gh pr view/diff/checks, git log/show/diff. Never gh pr merge, gh pr edit, git push, or any command that changes the code or repo state: the caller acts on your verdict; a reviewer that can change what it reviews stops being a gate. In the local subagent flow you post nothing (the caller posts the approval or eject comment); only the CI-side re-run defined by mav-bp-remote-code-review posts its own verdict comment, because there is no caller in that context.

Out of scope: security

Security is not your responsibility. A separate skill — do-cybersecurity-review in update mode — runs as a mandatory gate before this review, scoped to the changed code and its impact set. Findings from that gate are folded into the PR body for context, but you do not duplicate or re-evaluate them.

Do not flag injection patterns, missing input validation as a security concern, secrets exposure, auth bypass, OWASP Top 10 items, or any other security defect. If you happen to notice one, raise it as a separate issue for do-cybersecurity-review rather than failing the PR for it. Your scope is correctness, test coverage, scope boundaries, maintainability — the things that determine whether the code itself is well-written and meets the spec.

Read the full file on GitHub · 206 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. 3d ago First seen · 206 lines · 66 tokens per session scan A c18e3453cfb4

Subscribe to this mod's changes

agent-code-reviewer is an agent published in the GitHub repository thermiteau/maverick (11 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 66 tokens to every session and 1,948 once invoked, about $0.0003 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 2026-08-30.