puda-opentrons-vision-validation
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides specific bash instructions to capture images from an internal RTSP stream (rtsp://100.102.45.58:8554/cam0) using ffmpeg. It also utilizes standard file system utilities like mkdir, cp, stat, and sha256sum to manage the captured image files within the local environment.- [INDIRECT_PROMPT_INJECTION]: The skill processes external data (images and deck maps) which represents an attack surface for indirect prompt injection.
- Ingestion points: Deck images fetched via a local RTSP stream in SKILL.md and user-supplied deck configuration protocols.
- Boundary markers: The instructions include multiple directives for the agent to stop and request explicit user confirmation (e.g., 'Ask the user to confirm uncertain labware') and to avoid autonomous robot execution ('Do not execute robot motion until the visual check passes').
- Capability inventory: Access to the opentrons machine command set, network access via ffmpeg, and local file system operations (mkdir, cp, stat, sha256sum).
- Sanitization: The skill mandates a conservative inspection workflow and human-in-the-loop approval before any physical action is taken.
Audit Metadata