polars
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill focuses on local data processing using the Polars library, promoting reproducible practices such as version pinning and explicit assumption tracking.
- [SAFE]: Interactions with external services are restricted to vendor-owned SandBase APIs. The instructions mandate explicit user confirmation for paid calls and require the agent to present clear cost estimates.
- [SAFE]: The skill implements a data minimization strategy, instructing the agent to send only the minimum necessary data when interacting with external APIs.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies and manages the attack surface for indirect injection when processing external data. * Ingestion points: External datasets retrieved via sandbase_run and local files inspected during the workflow. * Boundary markers: The workflow requires restating constraints and acceptance checks to bound execution. * Capability inventory: Data analysis and transformation capabilities provided by the Polars library. * Sanitization: Quality gates are defined to validate data shapes, types, and units, alongside mandatory baseline comparisons.
Audit Metadata