documents
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user documents and external sources, creating a surface where embedded instructions in those files could attempt to influence the agent's behavior.
- Ingestion points: As defined in
SKILL.md, the agent reads conversation history, external sources, and existing documents inspected via$orbit:docs-artifact. - Boundary markers: The skill instructs the agent to "Separate the file to edit, a template to follow, a style example and sources of facts," which helps distinguish data sources but does not provide technical delimiters.
- Capability inventory: The skill utilizes
$orbit:docs-artifact,$orbit:sheets-artifact, and$orbit:writing-stylefor document creation, spreadsheet management, and style application. - Sanitization: The instructions require checking access permissions and following a
<confirmation_policy>before placing sensitive information into collaborative documents.
Audit Metadata