The Agent Skills Directory

[SAFE]: The skill provides a set of Python scripts and documentation for verifying academic citations using established services such as CrossRef, arXiv, and Semantic Scholar. These interactions are limited to retrieving bibliographic metadata and do not involve credential exposure or unauthorized data access.
[PROMPT_INJECTION]: The skill processes external bibliographic data (BibTeX and LaTeX files), which is a common vector for indirect prompt injection. 1. Ingestion points: Data is read from user-provided .bib and .tex files in scripts/verify-citations.py and scripts/format-checker.py. 2. Boundary markers: The scripts do not explicitly use delimiter-based isolation when processing file content for display in the agent context. 3. Capability inventory: The skill can perform network requests to academic APIs and write local verification reports. 4. Sanitization: Content is parsed using standard bibliographic libraries but is not filtered for natural language instructions. Given the skill's purpose, this risk is considered a known surface rather than an active threat.

citation-verification