The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's architecture defined in SKILL.md (Phase Φ2) describes the execution of several CLI tools, including pdf-search, pdf-brain, docling, and perplexity. These tools are integrated to perform semantic searches across PDF textbooks and convert documents into extractable sections for medical exam preparation.
[EXTERNAL_DOWNLOADS]: The skill utilizes the perplexity search tool to fetch latest research and current practice information from the web as part of its 'secondary enrichment' phase. This involves network requests to external servers to retrieve academic and clinical data.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a potential attack surface through the ingestion of untrusted data from external sources, specifically PDF textbooks (processed via docling and pdf-search) and web content (via perplexity).
Ingestion points: External PDF files and web search results.
Boundary markers: The skill uses a 'symbol lexicon' and strict citation formatting rules in references/citation-formats.md to structure outputs.
Capability inventory: The skill uses CLI tools for search, PDF conversion (docling), and a Python script for response validation.
Sanitization: The skill implements a 'Multi-Lens Critique' (Phase Φ6) and a systematic synthesis protocol (references/synthesis-protocol.md) that uses authority, recency, and consensus checks to validate and filter ingested data before it is presented to the user.

textbook-grounding