The Agent Skills Directory

[COMMAND_EXECUTION]: Spawns subprocesses to invoke the pandoc utility for manuscript rendering, bibliographic data parsing, and metadata stripping. Command construction uses standard argument parsing and avoids arbitrary shell execution.
[EXTERNAL_DOWNLOADS]: Retrieves bibliographic metadata and journal abbreviations from legitimate scientific services including the National Center for Biotechnology Information (NCBI) PubMed E-utilities (eutils.ncbi.nlm.nih.gov) and CrossRef (api.crossref.org). It also communicates with the local Zotero connector API on port 23119 for Zotero-integrated workflows.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied markdown files and BibTeX databases which represent an ingestion surface for untrusted content. Although the skill employs deterministic parsers and filters to handle this data, this surface could theoretically be leveraged to inject malicious instructions intended to influence downstream agent actions during the writing or peer-review stages.

manage-refs