workspace
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of data from untrusted sources into the agent's context, which could allow maliciously crafted files to influence agent behavior.
- Ingestion points: The
workspace_readtool is designed to access files in theuploads/directory, which contains user-provided content (SKILL.md). - Boundary markers: The documentation does not specify or require the use of delimiters or protective instructions when file content is interpolated into the agent's prompt context.
- Capability inventory: The skill includes tools to list, read, and write files within the session workspace, providing a mechanism for an agent to process and persist manipulated data (SKILL.md).
- Sanitization: There is no mention of content validation or sanitization for ingested files prior to model processing.
Audit Metadata