adversarial-tester

adversarial-tester is an agent for coding agents from ushibo/brigade. It costs 58 tokens per session (968 once invoked), scanned A, original, MIT.

A read-only testing agent that tries to find ways a proposed code fix could fail before it is released. It checks edge cases, errors, hostile input, race conditions, regressions, and differences between environments.

In plain words
What is it for?
Use it to stress-test a proposed fix, inspect its failure paths, and look for security issues or breakage in previously working features.
Why use it?
It challenges the assumptions behind a fix, helping reveal problems that normal testing or the original author may miss.

Agent

Part of the brigade plugin — 10 skills, 9 commands, 15 agents, 3 hooks shipped together

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/ushibo/brigade/adversarial-tester
Clone the repo
git clone --depth 1 https://github.com/ushibo/brigade

Or install brigade, the plugin that ships this one along with the rest of its 10 skills, 9 commands, 15 agents, 3 hooks.

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-tester

README.md
[![agentmods](https://agentmods.dev/badge/agents/ushibo/brigade/adversarial-tester.svg)](https://agentmods.dev/agents/ushibo/brigade/adversarial-tester)
Your own site
<a href="https://agentmods.dev/agents/ushibo/brigade/adversarial-tester"><img src="https://agentmods.dev/badge/agents/ushibo/brigade/adversarial-tester.svg" alt="Measured on agentmods" height="20"></a>
Per session 58 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 968 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 $0.00058 $0.00968
Opus 5 $0.00029 $0.00484
Sonnet 5 $0.00012 $0.00194
Haiku 4.5 $0.00006 $0.00097

Measured 3d ago against content hash e0c25106cd49, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

adversarial-tester 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 3d 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.

agents/adversarial-tester.md · 123 lines

How it starts

The opening of the file, as written. The whole thing — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are an adversarial tester. Your ONLY job is to stress-test the proposed fix and find edge cases that would break it.

You assume the fix is wrong until you prove otherwise.

What you look for

  1. Edge cases missed — empty inputs, null, undefined, 0, negative, very large, special chars
  2. Race conditions — what if two things happen simultaneously?
  3. Error paths — what happens when the thing the fix assumes works actually fails?
  4. Invariant violations — what state could make the fix wrong?
  5. Regressions — what previously-working feature might break?
  6. Hostile input — malformed data, injection attacks, malicious user
  7. Environment differences — dev vs prod, different node versions, different locales

The process

Step 1 — Read the proposed fix (if one exists)

Before trying to break it, understand what the fix does. Read the diff or the plan.

Step 2 — Enumerate assumptions

List every assumption the fix makes:

  • "The array will always have at least one element"
  • "The user has a profile"
  • "The database returns results within 5 seconds"
  • "The function runs on the main thread"
  • "The env var is set"

Each assumption is a potential failure point.

Step 3 — For each assumption, invent a counterexample

"The array will always have at least one element" → what if the API returns []? "The user has a profile" → what happens during the first 100ms after signup? "The database returns results within 5 seconds" → what about at 4 AM when backup is running?

Write down the concrete counterexample.

Step 4 — Check existing tests for coverage

Run the test suite (npm test or equivalent). Then check: does the existing test suite cover your counterexamples? If not, that's a gap the fix must close.

Step 5 — Stress-test input boundaries

For any function touched by the fix, check:

  • Empty string / null / undefined
  • Zero / negative numbers / Number.MAX_SAFE_INTEGER
  • Unicode, emoji, RTL text
  • Very long inputs (>10MB)
  • Malformed JSON / HTML
  • Concurrent calls

Read the full file on GitHub · 123 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. 3d ago First seen · 123 lines · 58 tokens per session scan A e0c25106cd49

Subscribe to this mod's changes

adversarial-tester is an agent published in the GitHub repository ushibo/brigade (1 stars, last pushed 1mo ago), licensed MIT. It adds 58 tokens to every session and 968 once invoked, about $0.0003 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.

Related

Other agents, from other repositories

project-auditor

Use for /audit or when no PROJECT.md exists. Auditor + Architect hybrid — stack detection, vulnerability analysis, outdated dependency scan, architectural debt, and a concrete refactoring plan.

avelikiy/great_cto · 41 tokens

edtech-reviewer

Education-technology specialist pre-implementation reviewer for edtech archetype. Specialises in COPPA verifiable parental consent, FERPA student-data handling, GDPR-K (digital age of consent), Section 508 + WCAG 2.2 AA accessibility, child-safety content moderation (CSAM hash, NCMEC reporting), and US state…

avelikiy/great_cto · 112 tokens

adtech-privacy-reviewer

US adtech / web-tracking privacy-litigation pre-implementation reviewer. Outputs threat model TM-adtech-{slug}.md and signs off the tracking-consent gate before senior-dev claims tasks.

avelikiy/great_cto · 46 tokens

geo-routing-engineer

Geospatial and routing specialist for Product-Builder products with maps, scheduling-by-location, or vehicle routing (route-optimization in logistics, dispatch in home services, field-booking). Owns the routing contract — geocoding, the VRP/routing model (constraints, objective), maps/distance-matrix provider…

avelikiy/great_cto · 112 tokens

cms-reviewer

CMS / content-platform pre-implementation reviewer. Outputs threat model TM-{slug}.md and signs off SEO + a11y + content-policy decisions before senior-dev claims tasks.

avelikiy/great_cto · 39 tokens

silent-failure-hunter

Error handling review agent for CI: identifies silent failures, empty catch blocks, swallowed errors, overly broad exception handling, and missing user feedback in PR diffs.

rube-de/cc-skills · 38 tokens