mmr-review

A command-line tool that sends a code change or design proposal to several AI reviewers, then combines their findings and can block completion based on severity. A design critique reviews an approach before implementation.

In plain words
What is it for?
Use it to review a GitHub pull request, staged changes, a branch difference, a file, or a design document.
Why use it?
Different reviewers may notice different correctness, security, or design problems, reducing reliance on one opinion before a merge or build.

Cursor rule

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 rules/zigrivers/scaffold/mmr-review
Clone the repo
git clone --depth 1 https://github.com/zigrivers/scaffold
Per session 78 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,161 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 $0.00078 $0.01161
Opus 5 $0.00039 $0.00580
Sonnet 5 $0.00016 $0.00232
Haiku 4.5 $0.00008 $0.00116

Measured 2d ago against content hash 5f85db9e4884, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

mmr-review 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.

packages/mmr/templates/skills/cursor/mmr-review.mdc · 103 lines

How it starts

The opening of the file, as written. The whole thing — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.

mmr — Multi-Model Review

Dispatch code reviews across several AI model CLIs (Claude, Codex, Grok and Antigravity by default; OpenCode opt-in), reconcile the findings, and gate on severity. Its peer mmr critique does the same fan-out for a design and is advisory (no gate).

Run a review

Pick the input mode that matches the target. Pass --sync --format json to get reconciled findings back in a single call:

# GitHub PR (fetches the diff via `gh pr diff`)
mmr review --pr <number> --focus "what to focus on" --sync --format json

# Staged changes (pre-commit)
mmr review --staged --sync --format json

# All tracked uncommitted changes (excludes untracked files)
git diff HEAD | mmr review --diff - --sync --format json

# Branch / ref range
mmr review --base main --head <branch> --sync --format json

# A specific file's current contents (tracked-no-changes, untracked, or new)
(diff -u /dev/null path/to/file.ts || true) | mmr review --diff - --sync --format json

The --diff flag expects diff-format content (a .patch/.diff path, or - for stdin). It does not read raw file content — wrap the target in a diff first. The || true guard is required because diff exits 1 when files differ, which breaks pipelines under set -o pipefail.

Severity gate

The verdict blocks on findings at or above fix_threshold (default P2; lower severities are advisory). Override per run with --fix-threshold P0|P1|P2|P3. Proceed only on pass or degraded-pass; fix blocking findings on blocked.

The verdict also reflects how many channels reported: fewer than defaults.min_completed_channels (default 2) completing yields needs-user-decision even with zero findings — one reviewer is not multi-model review. Treat that as "fix the channels" (mmr doctor), not as a pass.

Async flow (without --sync)

mmr review … prints a job id → mmr status <job-id> until complete → mmr results <job-id> --format markdown.

Avoid the nested self-review

Read the full file on GitHub · 103 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. 2d ago First seen · 103 lines · 78 tokens per session scan A 5f85db9e4884

Subscribe to this mod's changes

mmr-review is a cursor rule published in the GitHub repository zigrivers/scaffold (5 stars, last pushed 2d ago), licensed MIT. It adds 78 tokens to every session and 1,161 once invoked, about $0.0004 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.