The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to process external, untrusted machine learning papers, which presents an inherent surface for indirect prompt injection where a paper could contain hidden instructions to subvert agent behavior.
Ingestion points: Manuscript drafts provided by the user in SKILL.md.
Boundary markers: Absent. The instructions do not define specific delimiters or warnings to ignore embedded instructions within processed papers.
Capability inventory: WebSearch for finding citations and structured text generation.
Sanitization: No sanitization or validation of paper content is described.
[Data Exposure & Exfiltration] (SAFE): Network operations are limited to the WebSearch tool for identifying missing citations, which is consistent with the skill's stated purpose. No sensitive file paths, environment variables, or hardcoded credentials were detected.
[NO_CODE] (SAFE): The skill consists entirely of instructional markdown and does not include any executable scripts or package dependencies.

project-referee