Adversarial Verify

Adversarial Verify is a skill for Claude Code from ralfyishere/rules-with-receipts. It costs 123 tokens per session (1,493 once invoked), scanned A, original, MIT.

A review step that deliberately tries to find mistakes or weaknesses in an answer before it is delivered.

In plain words
What is it for?
Use it to challenge a completed result, test claims where possible, and check high-confidence answers for overlooked problems.
Why use it?
It reduces the chance of presenting an untested conclusion, especially for code changes, diagnoses, analyses, recommendations, and other important work.

Skill for Claude Code

Written for Claude Code: installed under .claude/.

Good fit Use it to challenge a completed result, test claims where possible, and check high-confidence answers for overlooked problems.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/ralfyishere/rules-with-receipts/adversarial-verify
Install

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.

Any agent
npx skills add ralfyishere/rules-with-receipts --skill adversarial-verify
Clone the repo
git clone --depth 1 https://github.com/ralfyishere/rules-with-receipts

Made for: Claude Code.

Wrote 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.

agentmods badge for Adversarial Verify

README.md
[![agentmods](https://agentmods.dev/badge/skills/ralfyishere/rules-with-receipts/adversarial-verify/github.svg)](https://agentmods.dev/skills/ralfyishere/rules-with-receipts/adversarial-verify)
Your own site
<a href="https://agentmods.dev/skills/ralfyishere/rules-with-receipts/adversarial-verify"><img src="https://agentmods.dev/badge/skills/ralfyishere/rules-with-receipts/adversarial-verify/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.

agentmods 80×15 button for Adversarial Verify

Your own site · 80×15
<a href="https://agentmods.dev/skills/ralfyishere/rules-with-receipts/adversarial-verify"><img src="https://agentmods.dev/badge/skills/ralfyishere/rules-with-receipts/adversarial-verify.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 123 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,493 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00123 $0.01493
Opus 5 $0.00062 $0.00746
Sonnet 5 $0.00025 $0.00299
Haiku 4.5 $0.00012 $0.00149

Measured 10d ago against content hash bfa745bd4e15, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

Adversarial Verify 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.

.claude/skills/adversarial-verify/SKILL.md · 81 lines

How it starts

The opening of the file, as written. The whole thing — 81 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial Verify

Purpose

Switch from author to attacker before the user does. Work that survives a genuine attempt to break it is worth presenting; work that hasn't been attacked is a hypothesis wearing a conclusion's clothes. This is a short, targeted pass — minutes, not a second project.

When to use this skill

  • Before presenting any High/Critical-tier deliverable (per effort-calibration).
  • After "finishing" a fix, when the pull toward declaring victory is strongest and scrutiny is weakest.
  • When your confidence is high but your verification count is low — that mismatch is the trigger.
  • Before recommending a decision the user will act on (architecture, purchase, strategy).

When NOT to use this skill

  • Low-tier tasks and simple factual answers — one sanity check suffices.
  • Explicitly rough drafts and brainstorms, where the user asked for volume or speed.
  • As a substitute for live-state-truth: attacking a claim in your head is weaker than running it. If a check is executable, execute it.

Operating procedure

Step 1 — Finish the draft first. Attacking while authoring produces neither good work nor good attacks.

Step 2 — Switch roles. You are now a skeptical senior reviewer whose job is to find the flaw. Assume there is one.

Step 3 — Run the attack list (pick the 3+ most relevant; don't run all mechanically):

Attack Question
Hidden assumption What am I treating as true that I never checked?
Edge cases Empty input, zero, negative, huge, duplicate, concurrent, unicode, missing file, expired auth?
Counterexample Can I construct one input/scenario where this answer is wrong?
Wrongness probe "What would have to be true for this to be wrong?" — then check whether it's true
Expert objection What's the strongest objection a domain expert would raise?
Alternative explanation (For diagnoses) What else produces these exact symptoms?
Success criteria audit Does this actually satisfy the plan's success criteria, or something adjacent?

Read the full file on GitHub · 81 lines

Changes

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.

  1. 10d ago First seen · 81 lines · 0 tokens per session scan A bfa745bd4e15

Subscribe to this mod's changes

Adversarial Verify is a skill published in the GitHub repository ralfyishere/rules-with-receipts (2 stars, last pushed 2mo ago), licensed MIT. It adds 123 tokens to every session and 1,493 once invoked, about $0.0006 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.