ai-red-teaming
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill includes an 'Adversarial Prompt Library' section containing examples of prompt injection techniques such as DAN, system prompt extraction, and instruction overrides. These are intended as test vectors for security assessments and are not instructions for the agent analyzing the skill.
- [INDIRECT_PROMPT_INJECTION]: The
ai_red_team_runner.pyscript possesses an attack surface for indirect injection by reading fromadversarial-prompts.yamland sending content to an LLM. Ingestion Point:adversarial-prompts.yaml. Boundary Markers: Structural separation using the OpenAI message object format. Capability Inventory: Network communication via theopenailibrary. Sanitization: None, as the script is designed to process raw adversarial content for evaluation. - [COMMAND_EXECUTION]: The documentation discusses shell command execution and tool abuse as test categories. However, the provided runner script does not execute shell commands; it uses
yaml.safe_loadfor safe data ingestion and restricts activity to API-based model interaction.
Audit Metadata