The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate_schematic.py uses subprocess.run to execute the local AI generation logic in scripts/generate_schematic_ai.py. The command is constructed as a list, which is a safe practice that prevents shell injection by avoiding shell interpolation of user arguments.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided descriptions to generate diagrams. This input is interpolated into prompts sent to external models (Nano Banana 2 and Gemini) via OpenRouter without sanitization or boundary markers.\n
Ingestion points: The prompt command-line argument in scripts/generate_schematic.py which is passed to the generation logic.\n
Boundary markers: Absent; the user input is directly embedded into the prompt templates in scripts/generate_schematic_ai.py.\n
Capability inventory: Network access (requests to OpenRouter) and file system access (writing PNG images and JSON review logs).\n
Sanitization: Absent; the user-provided string is interpolated without escaping or filtering.\n- [EXTERNAL_DOWNLOADS]: The skill makes network requests to openrouter.ai to generate images and obtain quality critiques. OpenRouter is a well-known platform for AI model access.

scientific-schematics