The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted data from user requests and Opentrons protocols to guide its execution.
Ingestion points: Expected deck maps and labware details are extracted from potentially untrusted user requests and Opentrons protocol files (SKILL.md).
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat embedded content within protocols as data rather than instructions.
Capability inventory: The skill utilizes the puda CLI and various shell commands (ffmpeg, mkdir, cp, stat, sha256sum) to manage images and communicate with hardware.
Sanitization: There is no instruction to validate or sanitize inputs such as labware names or filenames before interpolating them into shell commands.
[COMMAND_EXECUTION]: The skill uses shell utilities to facilitate its core functionality.
Evidence: Commands like ffmpeg for RTSP stream capture, mkdir, cp, and sha256sum for file management are used to handle robot deck images.

puda-opentrons-vision-validation