foucault-perspective
Pass
Audited by Gen Agent Trust Hub on Apr 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is purely instructional and educational, containing no executable logic or scripts. All provided content is in markdown and YAML format for persona definition and analytical modeling.
- [PROMPT_INJECTION]: The instructions contain directives for the agent to maintain a specific philosophical perspective (e.g., '拒绝本质主义' and '始终以福柯的思维框架回应'). These are benign persona-shaping instructions and do not contain attempts to override core safety protocols or reveal system prompts.
- [DATA_EXFILTRATION]: No network commands, sensitive file path access, or credential harvesting patterns were found. The skill operates solely within the user-agent conversation and does not request tools for external communication.
- [REMOTE_CODE_EXECUTION]: There are no remote script downloads or package installations. The installation command mentioned in the README is an informational guide referring to the author's own repository on a well-known platform.
- [COMMAND_EXECUTION]: The skill does not instruct the agent to execute shell commands, manage system processes, or perform any administrative actions.
Audit Metadata