explain-concepts
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to display visualizations using the 'feh' image viewer through shell commands.\n- [REMOTE_CODE_EXECUTION]: For mathematical and scientific topics, the skill generates and executes Python scripts using Matplotlib or Manim to create instructional diagrams.\n- [DATA_EXFILTRATION]: The skill records session summaries as markdown files within the 'learning-notes/' directory on the local file system.\n- [EXTERNAL_DOWNLOADS]: The skill performs web searches to find existing diagrams and uses external image generation tools for custom visuals when prompted.\n- [PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it processes untrusted user-defined concepts. Ingestion points: User concept requests in SKILL.md. Boundary markers: Absent. Capability inventory: Shell command execution (feh), dynamic Python execution, and local file system writes. Sanitization: Not specified.
Audit Metadata