The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external content such as research papers, drafts, and datasets. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the analyzed data to manipulate the agent's critique or final report.
Ingestion points: Processes user-provided drafts, data, and scientific papers as specified in the Agent Operating Procedure (SKILL.md).
Boundary markers: The skill relies on procedural instructions but lacks explicit delimiters or sandboxing for untrusted data processed at runtime.
Capability inventory: Access to Read, Write, and Edit tools, and the ability to execute a local python script for schematic generation.
Sanitization: No specific sanitization or filtering of input data is defined beyond standard agent safeguards.
[COMMAND_EXECUTION]: The skill includes instructions for the agent to execute a local python script (skills/scientific-schematics/scripts/generate_schematic.py) to generate visual aids. This execution is limited to a specific vendor-owned utility and is triggered only by explicit user request.
[EXTERNAL_DOWNLOADS]: The documentation describes optional integration with OpenRouter for schematic generation. It neutrally notes that prompts are sent to the third-party API and provides a disclosure regarding sensitive data transmission.

scientific-critical-thinking