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 patrickdappollonio/claude-plugins --skill adversarial-reviewgit clone --depth 1 https://github.com/patrickdappollonio/claude-pluginsWrote 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/patrickdappollonio/claude-plugins/adversarial-review)<a href="https://agentmods.dev/skills/patrickdappollonio/claude-plugins/adversarial-review"><img src="https://agentmods.dev/badge/skills/patrickdappollonio/claude-plugins/adversarial-review.svg" alt="Measured on agentmods" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Anti-Refusal · line 211 Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.Fix: Remove instructions that suppress warnings, disclaimers, or ethical commentary. Let the agent surface safety-relevant caveats to the user.
- medium Prompt Injection · line 50 Subtle instructions detected that may alter agent decision-making or introduce hidden biases.Fix: Review content for implicit steering or bias. Ensure instructions are explicit and align with the skill's stated purpose.
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 | $0.00213 | $0.13140 |
| Opus 5 | $0.00106 | $0.06570 |
| Sonnet 5 | $0.00043 | $0.02628 |
| Haiku 4.5 | $0.00021 | $0.01314 |
Grade A, and why
adversarial-review 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 3d 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.
Subtle steeringlowPrompt injection
Instructions that bias recommendations or shape behaviour without the user noticing.
Both come from a reviewer trusting a document instead of reasoning. The brief tells reviewers **what to look at**. It must never tell them **what to conclude**. 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 — 498 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Adversarial Review
Overview
Run a panel of independent, hostile reviewers against a change. Each reviewer assumes the code is broken and tries to prove it, from a single narrow angle. Because each runs as a fresh subagent, none of them inherit the main session's reasoning or the author's rationalizations — that is the whole point. A change that "looks fine" to the person who wrote it (or to the assistant that helped write it) gets attacked from 18 directions by reviewers who were never told why it should work.
This is the full panel, and it is token-heavy on purpose: 18 reviewers plus two gate agents, each reading the diff and the surrounding code. If the user asked for a smaller, quicker, lighter, or cheaper review, stop and run the adversarial-review-quick skill instead — it runs the 8 highest-yield of these reviewers (both design charters, tests, assumptions, observability, incomplete fixes, data integrity, API contract) with the same verifier and fix validator. Run this full panel when the user asks for a thorough or complete review, or when the change touches authentication, permissions, concurrency, untrusted input, failure-prone external dependencies, or anything that must roll back cleanly — those are the angles the quick panel drops.
Core principle: the orchestrator gathers the change and the brief once, hands each reviewer the raw change, the brief, and a charter, collects findings, a separate verifier confirms each finding is real, and a separate validator confirms each proposed fix actually works — before anything reaches the user.
The one thing reviewers must know: what was agreed
There are two very different things a reviewer could be told about a change, and they are easy to confuse:
| Give it to every reviewer | Never give it to any reviewer | |
|---|---|---|
| What it is | The brief — what this change was supposed to do, and what it was deliberately not going to do | Reassurance — anyone's opinion that the code is correct, safe, or already handled |
| Examples | The approved plan, mock, or design artifact; the issue or ticket; the PR description; the agreed non-goals and deferrals; agreed constraints ("don't touch the public API") | "The author says this is safe"; "this part is fine"; "focus on X, Y is handled"; your own hypotheses about where the bug is |
| Why | Without it, nobody checks the change against what was signed off, and every deliberate omission gets flagged as a gap — burying the real findings in noise | It tells the reviewer the answer before it looks, which is exactly the bias this skill exists to remove |
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.
- 3d ago Changed · +213 tokens per session 2a70c941e72c
- 8d ago First seen · 498 lines · 0 tokens per session scan A 56e66cf8cfe7
adversarial-review is a skill published in the GitHub repository patrickdappollonio/claude-plugins (8 stars, last pushed today), licensed MIT. It adds 213 tokens to every session and 13,140 once invoked, about $0.0011 per session on Opus 5. A static security scan graded it A with 1 finding (subtle steering). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
autoreview
Pre-commit/ship code review: Codex default; optional Claude or Pi.
rework-rate
Measure and interpret PR rework rate — the emerging 5th DORA metric.
omh-code-review
This is a Hermes-native code-review workflow skill.
revdiff-plan
Review the last Codex assistant message (plan, analysis, or proposal) with inline annotations in a TUI overlay. Extracts the most recent response from Codex rollout files and opens it in revdiff for review and annotation. Activates on "revdiff-plan", "review plan with revdiff", "annotate plan", "review last response"…
code-reviewer
Code review specialist focused on patterns, bugs, security, and performance.
agent-teams-simplify-and-harden
Implementation + audit loop using parallel agent teams with structured simplify, harden, and document passes. Spawns implementation agents to do the work, then audit agents to find complexity, security gaps, and spec deviations, then loops until code compiles cleanly, all tests pass, and auditors find zero issues or…