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 Eliyce/paqad-ai --skill adversarial-reviewgit clone --depth 1 https://github.com/Eliyce/paqad-aiWrote 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/eliyce/paqad-ai/adversarial-review)<a href="https://agentmods.dev/skills/eliyce/paqad-ai/adversarial-review"><img src="https://agentmods.dev/badge/skills/eliyce/paqad-ai/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/eliyce/paqad-ai/adversarial-review"><img src="https://agentmods.dev/badge/skills/eliyce/paqad-ai/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 | $0.00016 | $0.00617 |
| Opus 5 | $0.00008 | $0.00309 |
| Sonnet 5 | $0.00003 | $0.00123 |
| Haiku 4.5 | $0.00002 | $0.00062 |
Grade A, and why
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 12d 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 — 69 lines — stays where its author put it; the contents beside it link to each section on GitHub.
What It Does
Runs a risk-first review that looks for correctness defects, rollback hazards, missing coverage, and weak assumptions before work is treated as complete.
Use This When
Use this after design or implementation when the change is medium or high risk, customer-facing, security-sensitive, or hard to roll back safely.
Inputs
- Read the request, accepted plan, code or doc diffs, and any claimed verification results.
- Read the most relevant canonical docs for the changed behavior.
- Read
references/review-dimensions.mdbefore structuring findings.
Procedure
- Run
scripts/digest-evidence.shto flatten.paqad/session/verification-evidence.jsoninto agate | category | file:line | ac_id | messagetable; cite this table directly when anchoring findings. - Review the diff and digest against the dimensions in
references/review-dimensions.md— correctness, safety, performance, docs drift, verification sufficiency. - Prioritize user-visible failure, data corruption, migration risk, and contract breakage above style.
- Draft the output following
assets/output.template.md. Severity tags must come fromassets/severity-ranks.txt. Order findings non-increasing by severity. - Validate with
scripts/lint-findings.shbefore returning — exit 0 means the structural contract is met.
Output Contract
- Match
assets/output.template.md:## Findingsheading, one bullet per finding tagged**Critical|High|Medium|Low**, each citing concretefile:line(or gate name) and ending in aRequired action:segment. Optional## Open Questionssection. - Output must pass
scripts/lint-findings.sh(exit 0).
Escalate / Stop Conditions
- Block completion when a finding implies unsafe rollout, data loss, or broken customer behavior.
- Ask when the review depends on environment details or hidden rollout assumptions that are not present.
- Do not dilute confirmed defects with speculative style commentary.
Resources
references/review-dimensions.mdscripts/digest-evidence.shscripts/lint-findings.shassets/output.template.mdassets/severity-ranks.txtruntime/capabilities/coding/checklists/database-review-20pt.mdagents/openai.yaml
What ships with it
6 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.
- 12d ago First seen · 69 lines · 16 tokens per session scan A 11f2786947b2
adversarial-review is a skill published in the GitHub repository Eliyce/paqad-ai (8 stars, last pushed yesterday), licensed MIT. It adds 16 tokens to every session and 617 once invoked, about $0.0001 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
pr-writing-review
Extract and analyze writing improvements from GitHub PR review comments. Use when asked to show review feedback, style changes, or editorial improvements from a GitHub pull request URL. Handles both explicit suggestions and plain text feedback. Produces structured output comparing original phrasing with reviewer…
auto-review
A code-review skill that examines changed code using the TRUST 5 criteria and automated checks. It can review files, commit ranges, or changes assigned a risk level, with optional review strategies and providers.
ax-annotation
@AX code annotation workflow skill for agent-driven tag application.
ddd
A development method for improving existing code while preserving its current behavior. It follows an analyze, preserve, and improve cycle, using tests and small changes to reduce unintended effects.
review
A structured checklist for reviewing software code. It covers tests, readability, consistency, security, and whether changes can be tracked.
refactoring
A set of techniques for improving code structure while keeping its existing behavior unchanged. Refactoring means reorganizing code without intentionally changing what it does.