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.
npx skills add razaumair2203-ux/codex-adversarial-review-lite --skill codex-adversarial-review-litegit clone --depth 1 https://github.com/razaumair2203-ux/codex-adversarial-review-liteWrote 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.
[](https://agentmods.dev/skills/razaumair2203-ux/codex-adversarial-review-lite/codex-adversarial-review-lite)<a href="https://agentmods.dev/skills/razaumair2203-ux/codex-adversarial-review-lite/codex-adversarial-review-lite"><img src="https://agentmods.dev/badge/skills/razaumair2203-ux/codex-adversarial-review-lite/codex-adversarial-review-lite/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.
<a href="https://agentmods.dev/skills/razaumair2203-ux/codex-adversarial-review-lite/codex-adversarial-review-lite"><img src="https://agentmods.dev/badge/skills/razaumair2203-ux/codex-adversarial-review-lite/codex-adversarial-review-lite.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00076 | $0.18649 |
| Opus 5 | $0.00038 | $0.09324 |
| Sonnet 5 | $0.00015 | $0.03730 |
| Haiku 4.5 | $0.00008 | $0.01865 |
Grade A, and why
codex-adversarial-review-lite scanned grade A 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 11d 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.
Asks for rootlowPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
- `linux` / `wsl`: `sudo apt-get install -y git` if `apt-get` is available; otherwise `sudo dnf install -y git` if `dnf` is available; otherwise display manual guidance and stop. Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
How it starts
The opening of the file, as written. The whole thing — 1,470 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Codex Adversarial Review - Lite
Codex Adversarial Review - Lite is a Claude Code skill. It asks Codex CLI to independently audit Claude Code's plan, code changes, and focused test expectations before the builder trusts the work or applies fixes.
Tested setup:
- Builder: Claude Code
- Reviewer backend: Codex CLI
- Default reviewer model:
gpt-5.5 - Default reviewer reasoning:
xhigh
Cross-platform: Windows (Git Bash), macOS, Linux, and WSL with platform-aware sandbox defaults. Model fallback chain tries alternative models automatically if the default is unavailable. Self-test command validates the full tool chain without sending repo content.
The builder, reviewer, and review_backend terms keep the design portable later, but v1 is not a general multi-agent framework.
Invocation
Recommended v1 user invocation:
/codex-adversarial-review-lite audit- one reviewer pass, builder validates findings, report/HTML option is presented, user signs off before fixes./codex-adversarial-review-lite selftest- validate that all prerequisites, paths, model access, and platform behavior work on this machine. No repo content is sent. Run this first on any new machine.
Advanced scope hints:
/codex-adversarial-review-lite audit <file-path>- audit a specific file or plan./codex-adversarial-review-lite audit test-spec:<path>- audit with focused test expectations./codex-adversarial-review-lite audit test-data:<path>- audit with focused sample data or fixtures./codex-adversarial-review-lite audit rubric:<path>- audit against a domain checklist; the reviewer must report pass/fail per checklist item./codex-adversarial-review-lite audit strict rubric:<path>- high-consequence mode: requires a rubric, floor-gates every change for human review, disables autonomous fixing.
Options:
reviewer:<model>- defaultgpt-5.5.reasoning:low|medium|high|xhigh- defaultxhigh.sandbox:workspace-write|read-only|danger-full-access|inherit- default Unixworkspace-write, default Windowsdanger-full-access.approvals:auto_review|never- defaultauto_review. Both map to-c approval_policy=neverbecausecodex execis non-interactive; the sandbox flag is the real boundary. Interactive per-command approval (approvals:user) is not offered — there is no runtime prompt the user could answer during acodex execdispatch.test-spec:<path>- focused test expectations, scenarios, or validation commands to pass to the reviewer.test-data:<path>- sample inputs, fixtures, edge cases, or regression data to pass to the reviewer.rubric:<path>- domain checklist (markdown or plain text) injected into the reviewer prompt. The reviewer must report PASS/FAIL/UNVERIFIABLE per item in a# Rubric Resultssection. Any FAIL forcesVERDICT: REVISE.strict- high-consequence mode. Requiresrubric:<path>(stops with an error if missing), applies the human-review floor to every change regardless of category, and disables autonomous fixing even if the user previously asked for it.backend:codex- default and only implemented backend in v1.
What ships with it
2 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.
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.
- 11d ago First seen · 1,470 lines · 0 tokens per session scan A fd24c3382ace
codex-adversarial-review-lite is a skill published in the GitHub repository razaumair2203-ux/codex-adversarial-review-lite (1 stars, last pushed 2mo ago), licensed MIT. It adds 76 tokens to every session and 18,649 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 1 finding (asks for root). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
deep-review
Multi-angle code review that adapts to what is being reviewed — a mid-flight worktree increment, a full PR before production, a merge-window integration sweep, or a security-only pass. Findings are scored, adversarially refuted, widened to the same defect elsewhere in the codebase, and reported to the native review…
architect-review
Use when the user supplies an architecture artifact (assessment report, design doc, diagram, RFC, ADR) and asks for critique. Triggers on "review this", "what's wrong with", "is this any good", or any artifact-shaped paste with a question attached. Produces a verdict (SHIP IT / SHIP WITH CHANGES / MAJOR REWRITE /…
agent-review-panel
Orchestrate a multi-agent adversarial review panel: subagents with different perspectives independently review, debate and discuss, reach or fail consensus, then a supreme judge renders the verdict. Trigger on "review panel", "multi-agent review", "adversarial review", "panel review", "have agents debate this", "get…
Test Assertion Effectiveness 常に PASS するテストの検出
A test review that finds assertions—checks in tests—that do not really verify the program under test and therefore may always pass.
Knowledge-to-Code Alignment 新知識のコード反映・設計知識の保全
A code review check for whether new requirements and domain knowledge are reflected in names, responsibilities, and module boundaries without losing important design history.
Review Criteria Integrity レビュー基準・品質ゲートの自己弱体化検出
A review check for changes that weaken the rules or quality gates used to review the same change.