codex-review

codex-review is a command for Claude Code from adamjgmiller/adamsreview. It costs 37 tokens per session (3,487 once invoked), scanned B, original, MIT.

A Codex-based deep code-review command that produces the same report format as the regular Adams Review command. It uses Codex jobs for review stages such as finding and validating issues.

In plain words
What is it for?
Use it for a deep review of code changes when you want Codex to perform the review. Use the full option to force a complete review, or choose a higher effort level for more analysis.
Why use it?
It provides a Codex-only review path while keeping the report compatible with related review, fix, and walkthrough commands. Review effort can be set from low to extra high.

Command for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: reads .claude/ paths; mentions subagents; names the AskUserQuestion tool.

Part of the adamsreview plugin — 6 commands, 1 hook shipped together

Good fit Use it for a deep review of code changes when you want Codex to perform the review. Use the full option to force a complete review, or choose a higher effort level for more analysis.

Compare 6 commands from other repositories ↓
Install with agentmods
npx agentmods add commands/adamjgmiller/adamsreview/codex-review
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.

Clone the repo
git clone --depth 1 https://github.com/adamjgmiller/adamsreview

Made for: Claude Code.

Or install adamsreview, the plugin that ships this one along with the rest of its 6 commands, 1 hook.

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 codex-review

README.md
[![agentmods](https://agentmods.dev/badge/commands/adamjgmiller/adamsreview/codex-review/github.svg)](https://agentmods.dev/commands/adamjgmiller/adamsreview/codex-review)
Your own site
<a href="https://agentmods.dev/commands/adamjgmiller/adamsreview/codex-review"><img src="https://agentmods.dev/badge/commands/adamjgmiller/adamsreview/codex-review/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for codex-review

Your own site · 80×15
<a href="https://agentmods.dev/commands/adamjgmiller/adamsreview/codex-review"><img src="https://agentmods.dev/badge/commands/adamjgmiller/adamsreview/codex-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 37 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,487 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 1 finding. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00037 $0.03487
Opus 5 $0.00018 $0.01743
Sonnet 5 $0.00007 $0.00697
Haiku 4.5 $0.00004 $0.00349

Measured 10d ago against content hash 1cea10b0b2ed, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade B, and why

codex-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 10d 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.

CODEX_COMPANION="$(find ~/.claude/plugins -type f -name codex-companion.mjs -path '*codex*' 2>/dev/null | head -1)"
commands/codex-review.md · 270 lines

How it starts

The opening of the file, as written. The whole thing — 270 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Flags (optional):

  • --effort <level> controls Codex reasoning depth (low, medium, high, xhigh). Default high. Higher = deeper analysis but longer wall-clock and higher cost.
  • --full forces trivial_mode=false for this run (overrides the doc/config-PR early-exit). Same semantics as :review --full.

Note: this command has no --ensemble flag — it is purpose-built for Codex-only review. If you want internal Claude lenses pooled with the Codex CLI + a PR bot-comment scrape, run /adamsreview:review --ensemble instead.

Read fragments/_prelude-shared.md before proceeding — it lists rules that apply to every phase below (sub-agent return handling, helper-script error-as-prompt).

Execution overview

This command orchestrates the same phases as /adamsreview:review (Phases 0–6 plus Phase 5.5 auto-fix-hint generation), swapping the Claude sub-agent dispatches for Codex jobs (via the codex-companion.mjs plugin's task --background primitive) at:

  • Phase 1 detection — 7 parallel Codex lenses (L1–L7) instead of Claude Agent blocks.
  • Phase 4a deep validation — one parallel Codex job per surviving finding instead of one Opus sub-agent per finding.
  • Phase 4b light confirmation — chunked-batch Codex (≤25 candidates per chunk) instead of chunked-batch Sonnet.
  • Phase 5 cross-cutting — one Codex pass instead of one Opus pass.

Claude sub-agents (Sonnet) are still used for the "shape-fixer" / normalizer / dedup / scoring work where structured output matters more than reasoning depth — all of these phases execute the SAME fragments as :review:

  • Phase 0 preflightfragments/00-preflight.md verbatim.
  • Phase 2 dedupfragments/03-dedup.md verbatim.
  • Phase 3 scoring gatefragments/04-scoring-gate.md verbatim.
  • Phase 6 finalizefragments/07-finalize.md verbatim.

Each phase below is defined in a fragment under fragments/NN-<name>.md. At each phase boundary, read the named fragment with the Read tool and execute the instructions inside before proceeding to the next phase.

Before you start, build a TaskList that mirrors the phases below (one task per phase, plus one for argument parsing). Mark each in_progress when you start it and completed when you finish.

If a phase genuinely cannot run, mark the task completed with a one-line trace.md note and move on. Phase 6 (finalize) runs unconditionally — a partial review is still worth rendering and publishing so the user can inspect what did succeed.

Codex job failures (non-zero exit, malformed output, sentinel timeout) are handled per the adaptive retry-with-orchestrator-judgment policy: retry up to 3 times with the same prompt; on persistent failure, drop the affected unit (lens / finding / chunk) and escalate to the user via AskUserQuestion ("L2 failed; continue with 6 lenses or abort?"). Log every drop to trace.md with tag phase_<N>_codex_dropped:<unit_id>. This policy is restated where it applies in each Codex-using fragment.

Read the full file on GitHub · 270 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. 10d ago First seen · 270 lines · 37 tokens per session scan B 1cea10b0b2ed

Subscribe to this mod's changes

codex-review is a command published in the GitHub repository adamjgmiller/adamsreview (243 stars, last pushed 3mo ago), licensed MIT. It adds 37 tokens to every session and 3,487 once invoked, about $0.0002 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.