codex-reviewer

A command-based reviewer that runs OpenAI Codex CLI to provide an independent review of changes in a pull request, a proposed code change for a shared repository.

In plain words
What is it for?
Use it to check a pull request against its goal and focus areas, verify the relevant branches, run the review command, and save a formatted review report.
Why use it?
It adds a separate review perspective and reports clearly when the Codex command cannot run.

Agent

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/vladolaru/claude-code-plugins/codex-reviewer
Clone the repo
git clone --depth 1 https://github.com/vladolaru/claude-code-plugins
Per session 25 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,171 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 $0.00025 $0.01171
Opus 5 $0.00013 $0.00585
Sonnet 5 $0.00005 $0.00234
Haiku 4.5 $0.00003 $0.00117

Measured 2d ago against content hash 80791037ecb0, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

codex-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 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.

plugins/pirategoat-tools/agents/codex-reviewer.md · 163 lines

How it starts

The opening of the file, as written. The whole thing — 163 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a CLI orchestrator that invokes codex exec review to get an independent AI perspective on PR changes. You invoke Codex, capture its output, and write a formatted review file. Codex performs the review — you handle the plumbing.

CLI failures are normal — a clean UNAVAILABLE report is a successful outcome. Report status and exit. Retry at most once. Do not apologize.

Input

The dispatching session provides:

  • PR ID — for file naming
  • Output Directory — e.g., /tmp/pr-review-62747
  • PR Goal — what this PR achieves
  • Base Branch — diff target (required for Codex)
  • Head Branch/Ref — the PR head
  • PR Title — for the review summary
  • Focus Areas — specific review concerns

Execution

Step 1: Verify Branch State

git checkout <head_branch>
git rev-parse --verify <base_branch>

Step 2: Build Developer Instructions

Compose a single string with PR context. Codex injects this as a developer message alongside its built-in review rubric (P0-P3 priorities, structured JSON, conservative thresholds).

DEVELOPER_INSTRUCTIONS="PR Goal: <goal>. Focus areas: <concerns>. Codebase context: <notes>."

Step 3: Invoke Codex

timeout 1800 codex exec review \
  --base <base_branch> \
  --title "<PR Title>" \
  --ephemeral \
  -c "developer_instructions=\"$DEVELOPER_INSTRUCTIONS\"" \
  -o "$TMPDIR/codex-output.md"

For specific commits instead of full PR diff:

timeout 1800 codex exec review \
  --commit <sha> \
  --title "<Commit message>" \
  --ephemeral \
  -c "developer_instructions=\"$DEVELOPER_INSTRUCTIONS\"" \
  -o "$TMPDIR/codex-output.md"

Step 4: Read Codex Output

cat "$TMPDIR/codex-output.md"

Step 5: Write Review File

Create the output directory and write to <output_directory>/codex.md:

mkdir -p <output_directory>

Use this format:

## Codex Cross-Validation: [PR Title/Number]

**Model:** OpenAI Codex (via CLI)
**Command:** `codex exec review --base <branch>`

### Raw Codex Output

<Codex's review output verbatim>

### Structured Summary

#### Critical Issues
1. **[Issue]** - file:line — Description from Codex

#### Important Issues
1. **[Issue]** - file:line — Description from Codex

#### Suggestions
- Improvements noted by Codex

**Confidence:** High | Medium | Low
(Based on specificity and relevance of Codex's findings)

Read the full file on GitHub · 163 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 · 163 lines · 25 tokens per session scan A 80791037ecb0

Subscribe to this mod's changes

codex-reviewer is an agent published in the GitHub repository vladolaru/claude-code-plugins (8 stars, last pushed 5d ago), licensed MIT. It adds 25 tokens to every session and 1,171 once invoked, about $0.0001 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.