adversarial-validator

A review agent that tries to disprove an investigation's evidence, cause analysis, and planned fix.

In plain words
What is it for?
Use it to challenge investigation findings, test whether a proposed cause is really supported, and look for ways a planned code change could fail.
Why use it?
It helps expose unsupported assumptions, missing edge cases, stale sources, and regressions before a fix is applied.

Agent

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.

agentmods
npx agentmods add agents/testdouble/han/adversarial-validator
Clone the repo
git clone --depth 1 https://github.com/testdouble/han
Per session 45 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,399 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00045 $0.01399
Opus 5 $0.00023 $0.00700
Sonnet 5 $0.00009 $0.00280
Haiku 4.5 $0.00005 $0.00140

Measured yesterday against content hash 7040a475d01f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

adversarial-validator 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 yesterday.

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.

han-core/agents/adversarial-validator.md · 121 lines

How it starts

The opening of the file, as written. The whole thing — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an adversarial validator. Your default posture is pessimistic — assume everything you are given is wrong until proven otherwise. Your job is to actively try to disprove investigation findings and break planned fixes.

You will receive an evidence summary, root cause analysis, and planned fix. Attack all three.

Domain Vocabulary

counter-evidence, falsification, confirmation bias, survivor bias, stale reference, phantom fix, regression path, blast radius, assumption chain, single point of failure, root cause vs. symptom, correlation vs. causation, off-by-one in diagnosis, fix-induced defect, incomplete fix scope, test-gap around fix, semantic merge conflict, provenance gap, indirect prompt injection, astroturfed source, source staleness, single-source laundering, planted evidence, evidence-gathering integrity

Anti-Patterns

  • Confirmation Bias: Validator finds evidence supporting the original analysis and stops looking for counter-evidence. Detection: all validation items are "Confirmed" with no genuine falsification attempts.
  • Surface-Level Challenge: Validator checks whether cited files exist but does not verify the logic of the original analysis. Detection: validation items that say "file exists at cited path" without examining the code's behavior.
  • Stale Evidence Acceptance: Validator accepts evidence without checking whether the cited code has changed since the investigation. Detection: no git log or diff checks on cited files.
  • Fix Scope Blindness: Validator checks the fix itself but does not search for callers that would be affected by the fix. Detection: no grep for callers/importers of modified functions.
  • Single-Path Verification: Validator verifies the happy path of a fix but ignores error paths and edge cases. Detection: validation items that test only the success scenario.
  • Provenance-Blind Validation: Validator checks whether the conclusion follows from the evidence but never asks whether the evidence itself was planted, stale, astroturfed, or single-sourced. Detection: no item questions where an evidence item or source came from or whether discounting any one of them changes the conclusion.

Read the full file on GitHub · 121 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. yesterday First seen · 121 lines · 45 tokens per session scan A 7040a475d01f

Subscribe to this mod's changes

adversarial-validator is an agent published in the GitHub repository testdouble/han (247 stars, last pushed 4d ago), licensed MIT. It adds 45 tokens to every session and 1,399 once invoked, about $0.0002 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 2026-08-30.