Litrevu — agentic threat model

Not certain from the listing — likely utilizes commercial or open-source LLMs for text synthesis. The primary threat at this layer is indirect prompt injection, where instructions embedded in uploaded research papers could hijack the model's behavior during synthesis.

Handles user-uploaded research papers and articles. Key threats include data privacy leaks of unpublished manuscripts, data poisoning via malicious source documents, and vulnerabilities in PDF/document parsing libraries used to extract text.

Not certain from the listing — likely uses a basic RAG or document-chunking pipeline rather than a complex agentic framework. Risks include insecure handling of session state and potential context window overflow when processing large volumes of papers.

Not certain from the listing — deployment details are unspecified. If hosted as a freemium service, risks include insecure file storage for uploaded documents and lack of sandboxing for the document parsing environment.

Not certain from the listing — no observability or evaluation mechanisms are mentioned. Gaps here could allow hallucinated citations or biased synthesis to go undetected by the user.

Not certain from the listing — no compliance certifications (e.g., SOC2, GDPR) are mentioned. This is a significant gap if researchers upload proprietary or pre-publication intellectual property.

The agent operates as a standalone horizontal utility with no described multi-agent interactions or marketplace integrations, resulting in negligible ecosystem-level risk.