survey-paper
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources (ArXiv papers) using the
arxiv-mcp-servertools. It lacks specific boundary markers or instructions to ignore potential commands embedded within the papers, creating a surface for indirect prompt injection. - Ingestion points: Full paper content downloaded via
mcp__arxiv-mcp-server__download_paperandread_paper. - Boundary markers: Absent. The instructions do not tell the agent to distinguish between paper content and its own instructions.
- Capability inventory: The agent has the ability to write files to the local filesystem (
Obsidian 페이지 저장). - Sanitization: None specified for the content extracted from papers before being used to generate notes.
- [COMMAND_EXECUTION]: The skill instructs the agent to save files to a specific, hardcoded absolute directory path:
/Users/iamseungpil/LSP/study/Study/01_Papers/Survey/[Category]/. While this matches the provided author context (iamseungpil), hardcoding absolute paths to user home directories is a rigid practice that could lead to unauthorized file writes if the environment differs from the intended one.
Audit Metadata