loop-review

A read-only code-review command that makes several focused passes over one pull request or set of uncommitted changes. A pull request is a proposed code change submitted for review.

In plain words
What is it for?
Use it to inspect a pull request or local changes for coverage gaps, correctness problems, and other review issues, with evidence for each review area.
Why use it?
It reduces the chance that a reviewer fixates on one issue or repeats earlier findings while missing problems elsewhere. It reports findings and never edits, commits, pushes, or comments.

Command

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 commands/akholod/consensus-review/loop-review
Clone the repo
git clone --depth 1 https://github.com/akholod/consensus-review
Per session 43 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,238 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 $0.00043 $0.01238
Opus 5 $0.00022 $0.00619
Sonnet 5 $0.00009 $0.00248
Haiku 4.5 $0.00004 $0.00124

Measured yesterday against content hash 7e6412fef8c1, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

loop-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 yesterday.

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.

commands/loop-review.md · 85 lines

How it starts

The opening of the file, as written. The whole thing — 85 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Run the loop-review runtime and present its report.

What this is

/loop-review is the cheaper, sequential counterpart to /consensus-review. Where consensus-review runs a parallel panel of independent reviewers over the same change, loop-review builds the review context once and then makes several differentiated discovery passes over it, each told what is already covered and what is still open — but never told the previous pass's reasoning. That last part is the point: review models hyperfocus on one theme and miss what sits next to it, and persuasive context from an earlier pass makes the next one repeat it.

It reports; it never repairs

/loop-review finds problems and writes a report. It does not edit code, stage, commit, push, or comment on a pull request — and that is enforced, not merely intended:

  • every review pass is spawned with --disallowed-tools covering Edit, Write, NotebookEdit, Bash, Task, Read, Grep, Glob, WebFetch and WebSearch, so the reviewing model holds no tool capable of changing anything;
  • the runtime contains no mutating git or gh command. gh pr diff is the only forge call, and it reads;
  • its single write is the report file. Working files live in a temp directory outside the repo.

Applying fixes is a separate, human-initiated step. The intended cycle is: review → a human or a separate agent applies changes → review again on the new diff. The runtime recomputes the diff hash on every run, so a review of stale code cannot be mistaken for a review of the current one.

How to run it

node "${CLAUDE_PLUGIN_ROOT}/bin/loop-review.mjs" [flags]
Flag Meaning
--pr=<n|url> Review a pull request (reuses the gh path; the fallback fetch stays in a temp dir)
--range=<base..head> Review a git range
(neither) Review uncommitted changes against HEAD
--model=<m> Runner model (default sonnet)
--timeout=<sec> Per-pass wall clock (default 240)
--max-tokens=, --max-calls=, --max-ms= Hard budgets; exhausting one blocks CLEAN
--out=<path> Report destination (default .reviews/loop-review-<hash>.md)
--no-file Terminal only, no report written
--json Emit the whole run state as JSON on stdout instead of the report

Read the full file on GitHub · 85 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. yesterday First seen · 85 lines · 43 tokens per session scan A 7e6412fef8c1

Subscribe to this mod's changes

loop-review is a command published in the GitHub repository akholod/consensus-review (3 stars, last pushed 12d ago), licensed MIT. It adds 43 tokens to every session and 1,238 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.