feishu-drive
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests file names, folder titles, and metadata from the external Feishu cloud storage API. Maliciously named files or folders could potentially influence the agent's behavior if it interprets these names as instructions during directory browsing or file information retrieval.
- Ingestion points: The
feishu_drivetool returns dynamic data (names, tokens, types) from the Feishu cloud space inSKILL.md. - Boundary markers: No explicit delimiters or instructions to ignore embedded content are defined in the tool description.
- Capability inventory: The skill performs API operations on Feishu cloud storage; it does not exhibit shell command execution, local file system writes, or broader network capabilities beyond the Feishu service.
- Sanitization: No specific sanitization or validation logic for file/folder metadata is described.
Audit Metadata