prototype-pollution
Fail
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill documents dangerous code patterns that enable arbitrary command execution on a host system. Specifically, it provides exploitation gadgets for the EJS template engine and Node.js child_process module, using pollution of the Object prototype to inject system commands (e.g., 'process.mainModule.require("child_process").exec("COMMAND")').
- [PROMPT_INJECTION]: The skill uses 'AI LOAD INSTRUCTION' blocks to steer the behavior of the AI agent. These instructions command the agent to act as an expert attacker and to prioritize finding and developing high-impact prototype pollution exploits, including the construction of minimal RCE proofs-of-concept.
- [EXTERNAL_DOWNLOADS]: The skill references multiple external GitHub repositories hosting security research tools for prototype pollution. While intended for research, these tools reside outside the skill's direct control and include projects like 'pp-finder' and 'server-side-prototype-pollution'.
Recommendations
- AI detected serious security threats
Audit Metadata