ai-security - Assess Security Risks in AI and LLM Systems
Assesses AI/ML systems for prompt injection, jailbreaks, model inversion, data poisoning, agent tool abuse, and MITRE ATLAS mappings.
Tags
Updated: 2026-10-03Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Scan prompts for injection signatures
- Assess jailbreak resistance
- Score model inversion risk
- Score data poisoning risk
- Detect agent tool abuse
- Map findings to MITRE ATLAS
- Recommend guardrail controls
Inputs
- Test prompts
- Prompt test file
- Target type
- Access level
- Threat category scope
- Written authorization
Outputs
- Injection findings
- Risk scores
- Jailbreak assessment results
- MITRE ATLAS mappings
- Guardrail recommendations
- Process exit status
Requirements
- Python 3 runtime
- AI threat scanner tool
- Written authorization for gray-box or white-box testing