adversarial-ml-evasion

adversarial-ml-evasion is a skill for Claude Code from PurpleAILAB/Decepticon. It costs 36 tokens per session (2,285 once invoked), scanned A, original, Apache-2.0.

A guide to creating adversarial examples that make a trained machine-learning classifier produce the wrong result. These are deliberately altered inputs, such as images or files, rather than attacks on a chatbot's instructions.

In plain words
What is it for?
It covers understanding the model and crafting perturbations for systems used in image recognition, malware detection, intrusion detection, and spam filtering.
Why use it?
It helps security testers assess whether image, malware, intrusion, or spam classifiers can be fooled by small changes.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter.

Good fit It covers understanding the model and crafting perturbations for systems used in image recognition, malware detection, intrusion detection, and spam filtering.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/purpleailab/decepticon/adversarial-ml-evasion
About the project

Decepticon is an autonomous red-team agent that coordinates AI agents, security tools, sandboxes, and supporting services for authorized cybersecurity assessments. Security researchers and red teams can run it through its Docker stack, cloud service, command-line interface, or Python SDK, with the catalogue entries representing its available skills.

PurpleAILAB/Decepticon · 5,482 stars · on GitHub · decepticon.red

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 PurpleAILAB/Decepticon --skill adversarial-ml-evasion
Clone the repo
git clone --depth 1 https://github.com/PurpleAILAB/Decepticon

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-ml-evasion

README.md
[![agentmods](https://agentmods.dev/badge/skills/purpleailab/decepticon/adversarial-ml-evasion/github.svg)](https://agentmods.dev/skills/purpleailab/decepticon/adversarial-ml-evasion)
Your own site
<a href="https://agentmods.dev/skills/purpleailab/decepticon/adversarial-ml-evasion"><img src="https://agentmods.dev/badge/skills/purpleailab/decepticon/adversarial-ml-evasion/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-ml-evasion

Your own site · 80×15
<a href="https://agentmods.dev/skills/purpleailab/decepticon/adversarial-ml-evasion"><img src="https://agentmods.dev/badge/skills/purpleailab/decepticon/adversarial-ml-evasion.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 36 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,285 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector warn 7 Sept 2026
SkillSpector: 1 finding, up to medium

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • medium Data Exfiltration · line 45
    Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
    Fix: Verify the destination URL is trusted and necessary. Remove or replace with documented APIs. Ensure no secrets, tokens, or PII are transmitted.
How audits are shown
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 $0.00036 $0.02285
Opus 5 $0.00018 $0.01143
Sonnet 5 $0.00007 $0.00457
Haiku 4.5 $0.00004 $0.00229

Measured 7d ago against content hash 9c526e4ea8dd, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

adversarial-ml-evasion scanned grade A with 1 finding 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 7d 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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

curl -s -X POST "$TARGET/predict" \
packages/decepticon/decepticon/skills/standard/analyst/adversarial-ml-evasion/SKILL.md · 261 lines

How it starts

The opening of the file, as written. The whole thing — 261 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial ML Evasion

Adversarial examples are inputs modified with small, deliberate perturbations that reliably cause a trained classifier to produce incorrect predictions. The perturbation is typically imperceptible to humans or functionally irrelevant, while the model's decision boundary is crossed. This is distinct from LLM jailbreaks: the target is a trained discriminative model (CNN, gradient-boosted tree, SVM, LSTM) served via an API or embedded in a product.

Authorized use only. All testing must be conducted against systems you own or have explicit written permission to assess. Generating adversarial inputs against production ML APIs without authorization may violate the CFAA and equivalent statutes.


ATT&CK Mapping

Technique Use
T1562.001 — Impair Defenses: Disable or Modify Tools Evade ML-based AV/EDR/IDS
T1036 — Masquerading Make malicious content look benign to a classifier
T1027 — Obfuscated Files or Information Perturb a file/image to fool static ML analysis

1. Reconnaissance — understand the target model

Before crafting perturbations, determine the attack surface:

# 1a. Identify the ML stack from job postings, open-source repos, error messages
# Common deployment stacks: TensorFlow Serving, TorchServe, ONNX Runtime, SageMaker, AzureML

# 1b. Probe input shape and output format
curl -s -X POST "$TARGET/predict" \
  -H "Content-Type: application/json" \
  -d '{"data": [[0.0]*784]}' | jq .
# Note: number of output classes, confidence scores vs hard labels
# Confidence scores = white-box-equivalent gradient signal via finite differences

# 1c. Check if the API returns confidence values (score-based black-box) or label only (hard-label)
# Score-based: enables gradient estimation, ZOO, NES attacks
# Label-only: requires hard-label attacks (HopSkipJump, QEBA)

2. White-box evasion (full model access)

Use when you have the model weights (pentest scope, internal system, open-source model).

Read the full file on GitHub · 261 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. 7d ago First seen · 261 lines · 36 tokens per session scan A 9c526e4ea8dd

Subscribe to this mod's changes

adversarial-ml-evasion is a skill published in the GitHub repository PurpleAILAB/Decepticon (5,482 stars, last pushed 11d ago), licensed Apache-2.0. It adds 36 tokens to every session and 2,285 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.

Related

Other skills, from other repositories

langchain-patterns

Use when langChain/LangGraph patterns — chains, agents, tools, memory, retrieval, graph workflows. Use when working with langchain patterns.

oyi77/1ai-skills · 35 tokens

langchain

Framework for building LLM-powered applications with agents, chains, and RAG. Supports multiple providers (OpenAI, Anthropic, Google), 500+ integrations, ReAct agents, tool calling, memory management, and vector store retrieval. Use for building chatbots, question-answering systems, autonomous agents, or RAG…

synthetic-sciences/openscience · 79 tokens

ai-hacker

Use when aI and LLM security testing — prompt injection, model manipulation, data exfiltration via AI. Use when testing AI-powered applications, finding prompt injection vulnerabilities, or assessing LLM-integrated systems.

oyi77/1ai-skills · 47 tokens

langchain-architecture

Design LLM applications using LangChain 1.x and LangGraph for agents, memory, and tool integration. Use when building LangChain applications, implementing AI agents, or creating complex LLM workflows.

Harmeet10000/skills · 45 tokens

crawl4ai

Complete toolkit for web crawling and data extraction using Crawl4AI. This skill should be used when users need to scrape websites, extract structured data, handle JavaScript-heavy pages, crawl multiple URLs, or build automated web data pipelines. Includes optimized extraction patterns with schema generation for…

Harmeet10000/skills · 65 tokens

speech-to-text

Transcribe audio to text using ElevenLabs Scribe v2. Use when converting audio/video to text, generating subtitles, transcribing meetings, or processing spoken content.

Harmeet10000/skills · 37 tokens