The Agent Skills Directory

[SAFE]: The skill's primary functionality is to analyze research papers provided by the user. No malicious code, obfuscation, or unauthorized data access patterns were detected.
[EXTERNAL_DOWNLOADS]: The skill instructions suggest fetching venue-specific guidelines from external URLs or conference websites to improve the quality of the review. This behavior is documented and aligns with the skill's purpose.
[PROMPT_INJECTION]: The skill processes untrusted external content (PDFs, LaTeX source files, and guideline URLs) as part of its core workflow. This presents an indirect prompt injection surface, though it is used for analysis rather than executing commands.
Ingestion points: SKILL.md (Step 1: PDF and LaTeX reading; Step 0: Guideline URLs).
Boundary markers: The skill does not explicitly instruct the agent to ignore instructions embedded within the paper content, relying on the model's inherent ability to distinguish between data and instructions.
Capability inventory: Uses Read (file contents), Grep (pattern matching), and Glob (file listing). It does not include subprocess execution or network-writing capabilities.
Sanitization: No explicit sanitization or validation of the ingested paper content is defined.

research-paper-review