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 juanmhidalgo/claude-plugins --skill adversarial-doc-reviewgit clone --depth 1 https://github.com/juanmhidalgo/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/juanmhidalgo/claude-plugins/adversarial-doc-review)<a href="https://agentmods.dev/skills/juanmhidalgo/claude-plugins/adversarial-doc-review"><img src="https://agentmods.dev/badge/skills/juanmhidalgo/claude-plugins/adversarial-doc-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/juanmhidalgo/claude-plugins/adversarial-doc-review"><img src="https://agentmods.dev/badge/skills/juanmhidalgo/claude-plugins/adversarial-doc-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00153 | $0.03007 |
| Opus 5 | $0.00077 | $0.01503 |
| Sonnet 5 | $0.00031 | $0.00601 |
| Haiku 4.5 | $0.00015 | $0.00301 |
Grade A, and why
adversarial-doc-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 10d 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.
How it starts
The opening of the file, as written. The whole thing — 277 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Adversarial Doc Review
Dispatches a reviewer that hunts for the subset of documentation problems that actually matter: the ones that cause a bad architecture, design, or implementation decision. It is not a proofreader and not a completeness checker.
The reviewer is the discuss:doc-adversary subagent.
Why this is dispatched and not inlined
The review depends on the reviewer forming an independent model of the problem before absorbing the document's framing. That is impossible in a context window that already contains the document, its drafts, or the discussion that produced it — the framing is already there and the blind-read phase becomes theater.
Therefore: always run the reviewer in a fresh context. Never perform this review yourself in the main conversation.
If subagents are unavailable, tell the user plainly that the review will be weaker and why, and offer these fallbacks in order:
- Run the reviewer in a new session with only the doc path and the repo.
- Run it inline anyway, and mark the report as context-contaminated.
Do not silently downgrade. The contamination is the single largest failure mode of this skill.
Dispatch
Use the Agent tool with subagent_type: "discuss:doc-adversary". Do not
give it a name:, and do not deliberately background it.
Do not assume that gets you a blocking call that hands back the report inline.
Depending on the harness, an Agent call may run async no matter how you invoke
it, returning a task id and notifying you later. Never build the review
around the report coming back in the tool result. OUTPUT_PATH below is the
channel that works everywhere; treat anything the tool result happens to
contain as a bonus.
Pass the reviewer paths, never content. Do not summarize the document, do not paste excerpts, do not explain what it is trying to do, do not mention who wrote it or that it was generated. Every one of those transmits the framing you are trying to keep out.
The prompt is four lines and nothing else:
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.
- 10d ago First seen · 277 lines · 153 tokens per session scan A 40d8ac3acaa2
adversarial-doc-review is a skill published in the GitHub repository juanmhidalgo/claude-plugins (8 stars, last pushed 2d ago), licensed MIT. It adds 153 tokens to every session and 3,007 once invoked, about $0.0008 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
ADR Decision Quality
Ensure ADRs capture context, decision, alternatives, tradeoffs, and follow-ups in a way that prevents future drift.
Architecture Traceability & Consistency
Ensure design changes stay consistent across ADRs, diagrams, and specs; decisions are traceable; and drift is explicitly managed.
tech-spec
Technical Specification Document: Creates detailed technical specs (RFCs, design docs) for features, systems, and integrations. Use this skill whenever the user wants to write a tech spec, design doc, RFC, technical proposal, system design, or API specification. Also trigger when the user says 'spec out', 'design…
abusing-dpapi-for-credential-access
Extract and decrypt Windows DPAPI-protected secrets (Credential Manager, browser logins/cookies, Wi-Fi credentials, KeePass keys) online or offline using SharpDPAPI, SharpChrome, Mimikatz, or Impacket's dpapi.py, including domain-wide decryption via the DPAPI backup key. Use during authorized red-team…
discuss-red-team
CocoDiscuss Red-Team Mode — adversarial devil's advocate session that challenges spec assumptions before $plan. Invoked via $discuss --red-team.
openlore-brainstorm
Transform a feature idea into an annotated story using a Domain Sketch or Constrained Option Tree. Use when asked to brainstorm, explore, or shape a feature before implementation.