The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes Python scripts included within its own package (check.py and validate_refs.py) to validate the document structure and citation metadata. It also mentions standard system utilities like pdfimages or mutool for legitimate image extraction tasks.
[EXTERNAL_DOWNLOADS]: The skill makes HTTP requests to api.crossref.org to enrich and verify bibliographic references. Crossref is a well-known, authoritative service in the academic and research community. The skill's usage of the API follows standard practices, including rate limiting and identifying itself with a specific User-Agent.
[INDIRECT_PROMPT_INJECTION]: Because the skill ingests external document files (PDF, Word, or text), it is exposed to potential prompt injection via untrusted content. However, the skill instructions provide explicit defensive guidance, ordering the agent to ignore any directives embedded in the source text and to prioritize the removal of personal and confidential data during the import process.

proposal-import