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 Beaulewis1977/claude-plugin-for-codex --skill claude-adversarial-reviewgit clone --depth 1 https://github.com/Beaulewis1977/claude-plugin-for-codexWrote 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/beaulewis1977/claude-plugin-for-codex/claude-adversarial-review)<a href="https://agentmods.dev/skills/beaulewis1977/claude-plugin-for-codex/claude-adversarial-review"><img src="https://agentmods.dev/badge/skills/beaulewis1977/claude-plugin-for-codex/claude-adversarial-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.
<a href="https://agentmods.dev/skills/beaulewis1977/claude-plugin-for-codex/claude-adversarial-review"><img src="https://agentmods.dev/badge/skills/beaulewis1977/claude-plugin-for-codex/claude-adversarial-review.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 | $0.00051 | $0.00474 |
| Opus 5 | $0.00026 | $0.00237 |
| Sonnet 5 | $0.00010 | $0.00095 |
| Haiku 4.5 | $0.00005 | $0.00047 |
Grade A, and why
claude-adversarial-review 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 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.
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.
What it actually says
Claude Adversarial Review
Use a hostile-but-constructive review to surface failure modes that an ordinary review could miss.
Target and output
Default to current uncommitted changes or use the requested base ref with
--base. Pass a requested lens with --focus. Ask Claude to challenge
assumptions, alternatives, rollback, races, reliability, security, and user
impact. Require severity-ordered findings with file:line locations, followed
by residual risks and open questions. Never fix or apply findings.
When the user asks to continue the last compatible Claude session or challenge
its earlier conclusion, pass --resume. Use --fresh for an explicitly
independent critique.
Workflow
- Resolve
<plugin-root>/scripts/claude-companion.mjsto an absolute path. - Make exactly one runtime invocation. Runtime action:
adversarial-review. - Classify routine or bounded review as normal, substantial review as hard,
and the most complex or high-stakes review as very-hard. Pass the
classification with
--tier; explicit model and effort selections always win. Pass--cwdwith the current workspace and--jsonas separate arguments, plus selected target, focus, background, timeout, or session flags. - Write the free-form review task and context to stdin. Never place task text in an option, argument, or shell command. Use an argument array; never concatenate or interpolate values into a shell command. Use the execution tool's separate stdin channel; if it exposes an interactive process session, send the text there and explicitly close stdin before waiting.
- Keep this read-only. The runtime forces plan permission and safe-mode. Never edit, write, fix, or apply changes, and do not request bypass permissions.
- Attribute Claude's critique, verify it independently, and synthesize only after separating valid findings from speculation.
What ships with it
1 file 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 · 44 lines · 51 tokens per session scan A f545410b90ea
claude-adversarial-review is a skill published in the GitHub repository Beaulewis1977/claude-plugin-for-codex (3 stars, last pushed 1mo ago), licensed MIT. It adds 51 tokens to every session and 474 once invoked, about $0.0003 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.
Other skills, from other repositories
flow-next-impl-review
Carmack-level implementation review of changes via the configured backend. Use when asked to review code or a diff in a flow-next repo.
flow-next-spec-completion-review
Verify that a spec's completed tasks fully implement the spec requirements. Use at spec completion before close.
flow-next-plan-review
Carmack-level review of a flow-next spec or plan via the configured backend. Use when asked to review a plan or spec.
flow-next-resolve-pr
Resolve PR review feedback. Fetches unresolved threads, triages, fixes, replies and resolves via GraphQL. Use when asked to address review comments.
flow-next-worktree-kit
Manage git worktrees (create/list/switch/cleanup) and copy .env files. Use for parallel feature work, isolated review, clean workspace, or when user mentions worktrees.
pr
Comprehensive PR/issue review - analyzes architecture, tests, identifies unrelated changes mixed in, drafts review comment or issue comment. Use when user asks to review a PR, check a PR, look at PR changes, or comment on an issue.