The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Retrieves research papers and supplementary files from standard scholarly domains such as doi.org using curl commands.
[COMMAND_EXECUTION]: Utilizes shell utilities including curl for fetching remote assets and grep for searching through XML/text content of scientific articles.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface as it ingests and processes untrusted data from academic abstracts and full-text documents.
Ingestion points: External data enters the context through rp_search, rp_abstracts, and rp_fulltext tool outputs.
Boundary markers: The instructions do not define clear delimiters or markers to separate untrusted paper content from the agent's logic.
Capability inventory: The agent has access to file-writing tools, network operations via curl, and subprocess execution via grep.
Sanitization: No evidence of data sanitization or filtering of the retrieved research text before analysis.
[DYNAMIC_EXECUTION]: Recommends the "Progressive Enhancement Pattern" where the agent writes and executes its own Python scripts (e.g., screen_papers.py) to manage batch processing of literature.

evaluating-paper-relevance