academic-research

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). This is a direct install target pointing to an unverified GitHub repository (akillness/jeo-skills) used with npx, which can fetch and execute arbitrary code from an unknown user and is therefore potentially unsafe.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In the academic-pipeline Stage 2 literature sweep and Stage 2.5/5/6/7 execution, the agent ingests user-provided free-text inputs for the research topic, paper text, and reviewer comments, and then performs research/citation verification based on that content, so outsider-authored text can flow into the LLM without any prior selection of a specific trusted item by the agent.