scientific-schematics
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
Bashtool to run local Python scripts (scripts/generate_schematic.py) for diagram generation. - [PROMPT_INJECTION]: The skill includes deceptive metadata by referencing non-existent AI models ('Gemini 3 Pro').
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted natural language descriptions to drive AI-based code and image generation.
- Ingestion points: User-provided strings in CLI arguments.
- Boundary markers: None identified in the documented examples.
- Capability inventory: Broad access including
BashandWritetools. - Sanitization: No validation or sanitization of input strings is mentioned in the documentation.
Audit Metadata