dataroom-page
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources via tools like
runDatasetandlistComponents. While this introduces a potential attack surface if the connected data sources contain malicious payloads, the skill provides structured procedures for data mapping and warns against arbitrary script execution. - Ingestion points:
listComponents,getDatasetDetail,runDataset(fetching data from external systems). - Boundary markers: The skill does not explicitly instruct the agent to treat tool outputs as potentially untrusted data.
- Capability inventory:
newPage,updatePageConfig,insertDataset, andupdateDatasetallow writing configurations and data back to the DataRoom platform. - Sanitization: The instructions require manual conversion of flat fields to a JSON tree structure, which acts as a structural validation step.
- [DYNAMIC_EXECUTION]: The skill mentions a
scriptfield in component data binding for processing data (e.g.,bep.data). Although the documentation states this is executed at the front-end component level and is disabled by default for certain dataset types, it represents a dynamic code execution surface. - [COMMAND_EXECUTION]: The dataset functionality involves executing SQL queries and HTTP requests. The skill explicitly mitigates SQL injection risks by mandating the use of
#{}for parameter binding and prohibiting${}string interpolation.
Audit Metadata