The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied natural language to generate and review scientific diagrams, creating a surface where malicious instructions could potentially influence the AI's behavior.
Ingestion points: The prompt argument provided by the user via the command line, which is passed from the wrapper scripts/generate_schematic.py to the core logic in scripts/generate_schematic_ai.py.
Boundary markers: The AI prompt templates in scripts/generate_schematic_ai.py (specifically within generate_iterative and improve_prompt) use clear headers such as USER REQUEST: {user_prompt} to distinguish user input from system instructions.
Capability inventory: The skill's capabilities include executing local Python scripts via subprocess.run and performing network requests to the OpenRouter API (openrouter.ai) using the requests library.
Sanitization: There is no explicit sanitization or validation of the user-provided prompt before it is interpolated into the AI generation and review templates.
[COMMAND_EXECUTION]: The script scripts/generate_schematic.py uses subprocess.run to call scripts/generate_schematic_ai.py. This is a benign implementation of a CLI wrapper, as it executes a known locally distributed file using its absolute path relative to the current script and does not invoke a shell.
[EXTERNAL_DOWNLOADS]: The skill documentation and setup code reference the requests and python-dotenv libraries. These are well-known, standard Python packages for handling network communication and environment variables respectively, and their use is considered a safe practice.

scientific-schematics