skills.rest
pluginagentmarketplace
pluginagentmarketplaceCOMMUNITY
Explore all skills by this author →
GitHub

data-poisoning

Analyze training data for poisoning and backdoor attempts in ML workflows.

3|Updated Nov 18, 2025
Search Tags:#data-integrity#training-data#ml-security#adversarial-ai#poisoning-detection#data-poisoning#defense-testing
One-click install
npx skills add https://github.com/pluginagentmarketplace/custom-plugin-ai-red-teaming --skill data-poisoning
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: data-poisoning
Source: https://github.com/pluginagentmarketplace/custom-plugin-ai-red-teaming/tree/main/skills/data-poisoning
Command: npx skills add https://github.com/pluginagentmarketplace/custom-plugin-ai-red-teaming --skill data-poisoning

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill enables security teams and data scientists to validate AI models against training-data poisoning by simulating common poisoning techniques and assessing defense effectiveness in controlled environments.

Core Features & Use Cases

  • Poisoning Scenario Simulation: Test label flipping, backdoor injection, clean-label attacks, and LLM poisoning within a safe, reproducible workflow.
  • Detection & Assessment: Evaluate poisoning indicators, risk scores, and recommended mitigations across diverse datasets and training pipelines.
  • Defense Readiness & CI Integration: Integrate tests into CI/CD or data pipelines to continuously validate data integrity and model robustness.

Quick Start

Run the detector script to analyze a sample dataset. For example: python scripts/detect-poisoning.py. Review the resulting poisoning_report.json for risk guidance.

Frequently Asked Questions about data-poisoning

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I detect data poisoning in ML training datasets?▼

Detect data poisoning by running analysis scripts that evaluate datasets for label flipping, backdoor injection, clean-label attacks, and LLM poisoning. The tool outputs a poisoning_report.json containing risk scores and recommended mitigations.

What types of training data poisoning attacks can I test for?▼

You can test for label flipping, backdoor injection, clean-label attacks, and LLM poisoning. The tool simulates these scenarios to assess defense effectiveness and detect poisoning indicators across diverse datasets.

Can I integrate poisoning detection into my CI/CD pipeline?▼

Yes, you can integrate poisoning detection into CI/CD or data pipelines to continuously validate data integrity and model robustness. The tool supports input validation, structured output schemas, and extensible mappings.

How do I assess backdoor injection vulnerabilities in my training pipeline?▼

Assess backdoor injection vulnerabilities by simulating attacks within a reproducible workflow. The tool evaluates your training pipelines, generates poisoning indicators, and provides defense readiness scores.

Does this tool support testing LLM poisoning scenarios?▼

Yes, LLM poisoning is one of the labeled attack types supported. The tool analyzes training data to detect potential poisoning and backdoor attempts specifically targeting large language model workflows.

What format does the poisoning detection report use?▼

The poisoning detection report uses JSON format. Running the detector script generates a poisoning_report.json file containing risk guidance, poisoning indicators, and recommended mitigations.

Related Skills

Explore More Data & Analytics Skills→
maruakshaymaruakshay
4

fine-tuning-security

Detect data-poisoning and backdoor risks in fine-tuning pipelines.

Community
Advanced
jayjpatel9717jayjpatel9717

llm-security

Detect and mitigate LLM security vulnerabilities like prompt injection and RAG poisoning.

Community
Advanced
pradyumnasagarpradyumnasagar
6

agent-handoff

Automate inter-skill data transfer and provenance capture via JSON Material Passports.

Community
Advanced
GuitarAlchemistGuitarAlchemist

ix-adversarial

Evaluate machine learning model robustness against adversarial attacks like FGSM and PGD.

Official
Advanced