aig-agent-redteam
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill features a comprehensive library of attack payloads in the
modules/mutation-attack/operators/directory, including jailbreaks (e.g., DAN, STAN) and system-override prompts. These are intended for authorized security testing and use "canary markers" to verify safety boundaries. - [OBFUSCATION]: A utility script (
encodings.py) provides various methods to wrap test payloads, such as Base64, Unicode homoglyphs, and zero-width steganography. These techniques are used to simulate attempts to bypass safety filters during security audits. - [EXTERNAL_DOWNLOADS]: The skill can download security data from Tencent's official GitHub repository (
https://github.com/tencent/AI-Infra-Guard.git). This is handled via standard git commands and is a download from a well-known service by the same vendor. - [COMMAND_EXECUTION]: Local Python scripts and shell commands are used to perform infrastructure fingerprinting and static code analysis on user-defined targets.
- [DATA_EXPOSURE]: The infrastructure auditing module includes patterns to identify exposed configuration files and API keys in target systems as part of its scanning functionality.
Audit Metadata