pylabrobot
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes a cherry-picking workflow that processes external data from a user-supplied CSV file via
pd.read_csv(hit_list_csv). The contents of this file (source and destination wells) are used to direct physical or simulated hardware movements without explicit validation or boundary markers. - Ingestion points:
SKILL.md(Workflow 2) usespandasto read data from a CSV file. - Boundary markers: No delimiters or instructions are provided to the agent to treat the CSV data as untrusted.
- Capability inventory: The skill uses the
pylabrobotlibrary, which is capable of controlling physical robot hardware (Hamilton, Tecan, Opentrons) for aspiration, dispensing, and tip handling. - Sanitization: The implementation lacks sanitization or verification of the well addresses and volumes derived from the CSV before execution.
Audit Metadata