The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external research papers, PDF files, and bibliographic data which are untrusted data sources. While this presents a surface for indirect prompt injection, the risk is mitigated by the analytical nature of the skill and is inherent to its primary function of literature review.
Ingestion points: Processes .bib files and PDF directories during the 'Corpus Assembly' and 'Weakness Extraction' phases.
Boundary markers: No specific boundary markers are defined for the extracted literature content.
Capability inventory: Access to Read, Write, Edit, Bash(uv*), and Bash(paperpile*) tools.
Sanitization: No explicit sanitization or filtering of external content is described.
[COMMAND_EXECUTION]: The skill utilizes specialized bash environments (uv, paperpile) to manage Python dependencies and citation data. These tools are used for their intended purposes within the research workflow and do not show signs of malicious command construction.

weakness-scanner