explore
Pass
Audited by Gen Agent Trust Hub on Mar 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides behavioral instructions to establish a 'thinking partner' stance, which does not include malicious code, obfuscation, or unsafe command execution patterns.
- [COMMAND_EXECUTION]: The instructions include a strong negative constraint ('must NEVER write code or implement features') designed to prevent the agent from performing unauthorized modifications to the environment while in this mode.
- [DATA_EXPOSURE]: The skill allows the agent to read files and search the codebase to ground its discussions in reality; however, there are no network-facing capabilities or tools provided that would enable data exfiltration.
- [PROMPT_INJECTION]: The instructions use standard instructional language to define operational boundaries and do not contain patterns intended to bypass safety filters or ignore system constraints.
Audit Metadata