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.
git clone --depth 1 https://github.com/shawnclybor/clybor-claude-toolingWrote 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/commands/shawnclybor/clybor-claude-tooling/adversarial-review)<a href="https://agentmods.dev/commands/shawnclybor/clybor-claude-tooling/adversarial-review"><img src="https://agentmods.dev/badge/commands/shawnclybor/clybor-claude-tooling/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/commands/shawnclybor/clybor-claude-tooling/adversarial-review"><img src="https://agentmods.dev/badge/commands/shawnclybor/clybor-claude-tooling/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 | $0.00059 | $0.00478 |
| Opus 5.5 | $0.00024 | $0.00191 |
| Sonnet 5.5 | $0.00012 | $0.00096 |
| Haiku 4.5 | $0.00006 | $0.00048 |
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 1mo 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
/adversarial-review
Invoke the adversarial-review skill against a target.
Usage
/adversarial-review <target> [--round-note "why a 4th+ round is warranted"]
<target> is a file path (.claude/PRPs/foo/plan.md, src/api/auth.ts), or a pasted proposal.
What happens
- Load
rubric.mdbefore reading the target — must-fix is defined before anything is graded - Classify the target type; run the mechanical checker if one applies, and stop if it fails
- Drift pre-check — if the review has been ratcheting, surface it and ask before proceeding
- Spawn
chaos-engineer,simplifier,adversarial-reviewerin parallel, one question each - Synthesise: dedup, re-grade against the rubric, count self-inflicted findings separately
- Record the round, re-check drift, report a verdict
What makes it different from /quality-review
/quality-review spawns the same three agents with no rubric, no stop condition, and no memory of
previous rounds. That is fine for a one-off look at a proposal. Use this one when the target will be
reviewed more than once, because that is where review processes fail — the count ratchets, the
bar rises silently, and findings the last round introduced get reported as the target degrading.
Do NOT auto-apply fixes
The skill reports. The user decides what lands. It never edits the target.
When to skip
- A routine code diff →
/code-review - A build plan's pre-execution gate with a profile and mirror →
build-validate - Trivial changes, or questions rather than proposals
- The stop condition already fired. Three rounds with the checker green and zero must-fix means review is done. Build the thing.
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.
- 1mo ago First seen · 44 lines · 59 tokens per session scan A 3325d80f84dc
adversarial-review is a command published in the GitHub repository shawnclybor/clybor-claude-tooling (6 stars, last pushed 6d ago), licensed MIT. It adds 59 tokens to every session and 478 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-05.
Other commands, from other repositories
audit
Review past Claude Code sessions for PBR workflow compliance and UX quality.
code-audit
Adversarial code review tree. Wraps /cc-tree:tree with the code-audit preset preselected — finds security / performance / correctness / contract findings a static linter would miss. Every finding carries file:line evidence + threat-model context + proposedfix; hard-banned against NEEDS-MORE-INFO deferrals.…
review-csk
Review pass — review + security + quality gates.
sonar
A command for running SonarQube, a tool that examines source code for quality and security problems. It can show current results, rerun the analysis, inspect security hotspots, check test coverage, and explain quality-gate results.
mise-en-place
Autonomous backlog run — flight-recorder prep, then run-through.
distill
Distill session artifacts to wiki and decisions; archive specs, drop scratch.