The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill communicates with several well-known academic indices including Semantic Scholar, Crossref, OpenAlex, and arXiv to verify research citations and retrieve metadata. These communications are directed toward official API endpoints for established academic services.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface because it ingests untrusted data from web search results and external PDF files. This is addressed by a mandatory evidence chain in the instructions: (1) Ingestion happens in bibliography and source verification agents. (2) Explicit 'instruction-data-boundary' rules are defined to prevent the agent from obeying commands embedded in data. (3) Capabilities are restricted to specific network APIs and directory-scoped file operations. (4) The skill includes self-reflection and quality gates to monitor output integrity.
[COMMAND_EXECUTION]: The skill utilizes the 'pdftotext' command-line utility within the timeline extraction agent to scan the first page of PDF documents for publication dates. This is a standard utility used for its intended purpose within the research workflow.

deep-research