The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest large amounts of untrusted content from external sources (academic papers, abstracts, and full-text XML/HTML) and process them to score relevance and extract data.
Ingestion points: Fetches content from ncbi.nlm.nih.gov (PubMed/PMC), api.unpaywall.org, biorxiv.org, and various publisher domains via DOI resolution.
Boundary markers: The instructions do not specify the use of delimiters (e.g., XML tags or special tokens) to separate external paper content from the agent's internal reasoning or tracking logic.
Capability inventory: The skill uses curl for network requests, performs local file writes to track progress (papers-reviewed.json, SUMMARY.md), and executes shell commands like grep for text analysis.
Sanitization: There are no explicit instructions to sanitize or escape the content retrieved from external papers before it is processed by the agent or written to tracking files.
[DYNAMIC_EXECUTION]: The skill encourages the agent to generate and execute local Python scripts (screen_papers.py, deep_dive_papers.py) for handling large datasets.
Evidence: The "Helper Scripts" and "Progressive Enhancement Pattern" sections provide templates and logic for the agent to write and run its own automation scripts.
Risk: While the logic is based on skill templates, generating and running executable code at runtime is a dynamic execution pattern.
[EXTERNAL_DOWNLOADS]: The skill performs numerous network operations to fetch research data.
Evidence: Uses curl to interact with ebi.ac.uk (ChEMBL), ncbi.nlm.nih.gov (PubMed), and unpaywall.org.
Context: These are all well-known, established scientific and academic services used for their intended purpose in this research-focused skill.

Evaluating Paper Relevance