codex-review-loop

codex-review-loop is a skill for Claude Code, Codex from choi138/toki. It costs 98 tokens per session (2,203 once invoked), scanned A, original, MIT.

A local code-review workflow for examining uncommitted changes, branches, or commits through specialist review paths. It can also apply fixes that the user explicitly approves.

In plain words
What is it for?
Use it for local code reviews, review-and-fix passes, re-reviews, and approved repairs to reported findings.
Why use it?
It helps organize review findings consistently and prevents edits from being made without clear approval.

Skill for Claude CodeCodex

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 skills/choi138/toki/codex-review-loop
Any agent
npx skills add choi138/toki --skill codex-review-loop
Clone the repo
git clone --depth 1 https://github.com/choi138/toki

Made for: Claude Code, Codex.

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-loop

README.md
[![agentmods](https://agentmods.dev/badge/skills/choi138/toki/codex-review-loop.svg)](https://agentmods.dev/skills/choi138/toki/codex-review-loop)
Your own site
<a href="https://agentmods.dev/skills/choi138/toki/codex-review-loop"><img src="https://agentmods.dev/badge/skills/choi138/toki/codex-review-loop.svg" alt="Measured on agentmods" height="20"></a>
Per session 98 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,203 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 $0.00098 $0.02203
Opus 5 $0.00049 $0.01102
Sonnet 5 $0.00020 $0.00441
Haiku 4.5 $0.00010 $0.00220

Measured 3d ago against content hash c82467ea62e2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

codex-review-loop 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 3d ago.

The scan reads SKILL.md. This mod also ships 18 executable files (scripts/finding_grouping.py, scripts/finding_validation.py, scripts/lane_registry_validation.py, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

.agents/skills/codex-review-loop/SKILL.md · 226 lines

How it starts

The opening of the file, as written. The whole thing — 226 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Toki Codex Review Loop

Run an rvw-compatible lane review, then apply the codex-lb-style approval, atomic-fix, verification, and bounded re-review loop.

Core Contract

  • Treat every review request as read-only until the user explicitly approves a write mode.
  • Never infer permission to edit or commit from phrases such as "review this" or from a GitHub @codex review request.
  • Never push, submit a GitHub review, comment, resolve threads, label, merge, or otherwise mutate remote state from this skill.
  • Preserve unrelated staged, unstaged, and untracked changes. Never use git reset, git checkout, or broad restoration to undo a finding fix.
  • Treat local usage logs, databases, audit findings, prompts, transcripts, and credentials as sensitive. Do not print raw matched diff content.
  • Apply project-conventions and read its task-specific references before fixing Toki source, tests, resources, or project configuration.
  • Apply .agents/conventions/git-workflow.md before any commit. A commit requires explicit user authorization separate from review authorization.

Trust Boundary

Scope resolution runs before any review and must survive a hostile repository. Treat this section as the contract a finding is measured against: report a finding when it breaks an invariant below, and classify it wont_fix with this section as the reason when it only restates a non-goal.

Untrusted Inputs

  • Repository file contents and tree layout, including untracked and ignored paths.
  • Repository Git configuration: .git/config, .gitattributes, .gitmodules, and any configured hook or helper program.
  • The inherited PATH and every GIT_* environment variable.
  • Lane result JSON, including every path and line number it reports.

Enforced Invariants

  1. Only trusted executables run. git and the review binary resolve from the platform default path, never from an inherited PATH, and the runner narrows PATH before invoking any external command.
  2. No repository-configured program executes. Content filters, textconv, external diff, core.fsmonitor, core.hooksPath, and lazy fetch are overridden to inert values or the scope is refused.
  3. Workspace confinement is derived from the filesystem, not from Git. The inventory walks the real tree from the repository root and validates every symbolic link target, so ignore rules and index modes cannot hide a path.
  4. Every scope fails closed. A submodule, embedded repository, escaping symlink, non-UTF-8 path, or unbounded inventory marks the scope unsafe and stops before Codex is invoked.
  5. Paths keep byte fidelity across process boundaries. Roots and path lists cross as NUL-terminated data and are never trimmed.
  6. Reported findings are untrusted data. A finding path must be repository-relative, free of NUL, and free of empty, current, or parent segments before it is used.
  7. The baseline lane is always on. A registry that marks it otherwise is rejected.

Read the full file on GitHub · 226 lines

Files

What ships with it

31 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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 · 226 lines · 98 tokens per session scan A c82467ea62e2

Subscribe to this mod's changes

codex-review-loop is a skill published in the GitHub repository choi138/toki (9 stars, last pushed 3d ago), licensed MIT. It adds 98 tokens to every session and 2,203 once invoked, about $0.0005 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.

Related

Other skills, from other repositories