scientific-schematics
Pass
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). User-supplied diagram descriptions are interpolated into system prompts for both image generation and quality review without significant isolation or sanitization.\n
- Ingestion points: User input from the
promptcommand-line argument inscripts/generate_schematic.py.\n - Boundary markers: Minimal; the prompts use simple labels like
USER REQUEST: {user_prompt}but lack hardened delimiters or instructions to the model to ignore embedded commands.\n - Capability inventory: The skill can perform network requests to the OpenRouter API and execute local helper scripts via
subprocess.run.\n - Sanitization: No sanitization or escaping of the user-supplied prompt string was found.\n- [COMMAND_EXECUTION]: The script
scripts/generate_schematic.pyuses thesubprocessmodule to execute the AI generation logic in a child process.\n - Evidence: The command is executed using a list of arguments and without a shell (
shell=Falseby default), which protects against shell injection. Furthermore, sensitive credentials (the API key) are passed via the process environment rather than command-line arguments, preventing them from appearing in system process listings.
Audit Metadata