The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user queries and external research content, which creates a potential surface for indirect prompt injection.\n
Ingestion points: Research queries provided via command-line arguments and results returned from the OpenRouter API.\n
Boundary markers: Queries are delimited by double quotes within the LLM prompt template in research_lookup.py.\n
Capability inventory: The skill is granted Bash, Read, Write, and Edit tool permissions in its configuration.\n
Sanitization: No explicit sanitization or filtering of input queries or retrieved research data is implemented.\n- [EXTERNAL_DOWNLOADS]: The skill fetches research data and model information from OpenRouter, a well-known service.\n
Evidence: Makes network POST and GET requests to https://openrouter.ai/api/v1 to interact with Perplexity Sonar models.\n- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute local scripts using the Bash tool for workflow automation.\n
Evidence: SKILL.md provides examples for running python scripts/generate_schematic.py and python lookup.py to process research queries and generate diagrams.

research-lookup