The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content retrieved from external academic databases, which is a potential surface for indirect prompt injection if malicious instructions were embedded in source materials.
Ingestion points: The workflow in SKILL.md uses web_search and web_fetch to retrieve external content from scholarly platforms.
Boundary markers: The skill does not define explicit delimiters or 'ignore' instructions for the fetched data, though it emphasizes critical synthesis and source verification.
Capability inventory: The skill utilizes web_search, web_fetch, and can execute Python scripts to process and format the document.
Sanitization: The skill relies on manual 'Source Verification' protocols but lacks automated sanitization for prompt injection patterns in retrieved text.
[EXTERNAL_DOWNLOADS]: The skill makes legitimate requests to well-known academic services to fetch research data.
Evidence: Referenced services include Google Scholar (scholar.google.com), IEEE Xplore (ieeexplore.ieee.org), PubMed (pubmed.ncbi.nlm.nih.gov), ACM Digital Library (dl.acm.org), and arXiv (arxiv.org).
[COMMAND_EXECUTION]: The skill includes a specialized Python utility for managing reference formatting.
Evidence: The file scripts/ieee_formatter.py provides logic for the agent to format IEEE citations for various publication types programmatically.

academic-research-writer