ai-redteam
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill provides a structured framework and documentation for security professionals to conduct authorized red teaming of AI applications, adhering to ethical guidelines and rules of engagement.
- [PROMPT_INJECTION]: The skill identifies prompt injection as a threat class and lists common techniques like refusal suppression and instructions override, but these are provided as reference material for testing rather than active injection attempts.
- [OBFUSCATION]: While the attack catalog mentions encoding techniques such as Base64 as potential jailbreak vectors to be tested, no actual obfuscated code, hidden instructions, or encoded URLs are present in the skill files.
- [DYNAMIC_EXECUTION]: The documentation includes illustrative pseudocode for designing an automated testing harness, which serves as a conceptual guide and does not include executable scripts or dynamic runtime compilation.
- [NO_CODE]: The skill contains no executable code or scripts, consisting entirely of markdown instructions and reference templates.
Audit Metadata