code-review

A command that sends code to several AI models for independent review, compares their findings, and combines them through structured agreement rounds.

In plain words
What is it for?
Use it to inspect code for defects, security concerns, or design problems and receive a unified review from multiple configured models.
Why use it?
It reduces dependence on one model's perspective and preserves progress during a long review. It saves intermediate results so the review can resume after context is lost.

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/altimateai/claude-consensus/code-review
Clone the repo
git clone --depth 1 https://github.com/AltimateAI/claude-consensus
Per session 18 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 6,239 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 1 finding. 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 $0.00018 $0.06239
Opus 5 $0.00009 $0.03120
Sonnet 5 $0.00004 $0.01248
Haiku 4.5 $0.00002 $0.00624

Measured 3d ago against content hash 3e6ee26cfc1a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

code-review scanned grade B with 1 finding 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.

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

[ -f ~/.claude/.env ] && export OPENROUTER_API_KEY=$(grep '^OPENROUTER_API_KEY=' ~/.claude/.env | cut -d= -f2- | tr -d '"')
plugins/consensus/commands/code-review.md · 513 lines

How it starts

The opening of the file, as written. The whole thing — 513 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Multi-Model Code Review

Get independent code reviews from multiple AI models (Claude + configured external models), compare findings, and converge on a unified review through structured synthesis and agreement rounds.

Compaction Resilience

This is a long-running command. Context compaction may erase in-memory state mid-run.

  • Check resume: Glob data/scratch/active-progress-consensus-code-review-*.md — find any with Status IN_PROGRESS and < 24h old. If found, read its SESSION_DIR and skip to first unchecked goal. If Status is COMPLETED or FAILED, ignore it.
  • Write progress: After creating SESSION_DIR, create a unique progress file: data/scratch/active-progress-consensus-code-review-{SESSION_ID}.md (where SESSION_ID is the random suffix from SESSION_DIR, e.g. X4f2kL)
  • Mark done: Set Status to COMPLETED at end of successful run, or FAILED on abort
  • Save incrementally: Write/append to $SESSION_DIR files after each phase, not at the end

Goals Template

# consensus:code-review — {TARGET}
Started: [timestamp]
Status: IN_PROGRESS
Command: consensus:code-review
SESSION_DIR: {SESSION_DIR path}
TTL: 24h

## Goals
- [ ] Phase 1 — Setup: load config, determine review target, write prompt, create team
- [ ] Phase 2 — Spawn reviewers: launch teammate agents + write Claude's code review
- [ ] Phase 3 — Collect reviews: wait for all models to send their reviews
- [ ] Phase 4 — Analyze & compare: build comparison table, identify consensus issues
- [ ] Phase 5 — Synthesize: draft unified code review ordered by severity
- [ ] Phase 6 — Convergence: send draft to all models, collect APPROVE/CHANGES NEEDED
- [ ] Phase 7 — Present final review with attribution table, cleanup team

## Progress
- [HH:MM] Starting execution...

Checkpoints

  • After Phase 2: Claude's review written to $SESSION_DIR/claude.md
  • After Phase 3: All model reviews on disk at $SESSION_DIR/{model.id}.md
  • After Phase 5: Draft review at $SESSION_DIR/draft.md
  • After Phase 6: Convergence responses at $SESSION_DIR/{model.id}-convergence.md

Read the full file on GitHub · 513 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 · 513 lines · 18 tokens per session scan B 3e6ee26cfc1a

Subscribe to this mod's changes

code-review is a command published in the GitHub repository AltimateAI/claude-consensus (32 stars, last pushed 8d ago), licensed MIT. It adds 18 tokens to every session and 6,239 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.