The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill ingests untrusted external data in the form of academic papers as noted in the 'When to Apply' section of SKILL.md. No boundary markers or sanitization logic are defined in the instructions to separate paper content from system instructions. However, the capability inventory reveals zero risky operations (no shell, file, or network access across the skill), restricting the threat to reasoning influence or output poisoning.
[No Code] (SAFE): No executable code, scripts, or package manifests were found. The skill operates solely through natural language instructions.
[Data Exposure] (SAFE): No sensitive file paths, credentials, or network operations were detected in the content.

academic-researcher