The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill performs network requests to scholarly metadata services including api.crossref.org, api.openalex.org, and doi.org to fetch publication data and resolve identifiers. These are well-known infrastructure services for academic research.
[DATA_EXFILTRATION]: The helper scripts in kernel.py are designed to retrieve the user's email address via platform-provided tools (host.get_user_email()) and include it in the User-Agent header or URL parameters of requests sent to Crossref and OpenAlex. This behavior is explicitly disclosed in the skill's YAML frontmatter as a mechanism to access the 'polite pool' for these APIs, which provides more reliable service for scholarly tools.
[COMMAND_EXECUTION]: The instructions in SKILL.md provide a fallback mechanism for the agent to read and execute (exec()) its own kernel.py file if its helper functions are not automatically loaded. This is a local execution of the skill's own provided source code.
[DATA_EXPOSURE]: The skill processes external scholarly data (titles, abstracts, and citations) which is interpolated into the agent's output. While this presents a potential surface for indirect prompt injection if a third-party source contained malicious instructions, the skill includes a style_pass linting function that reviews the final draft for specific phrases and formatting issues.

literature-review