The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.Popen within its threading modules (thread.py) to invoke a local utility named puda. This command is used to monitor real-time telemetry from a mass balance during experiments. The execution uses list-based arguments and is a core functional component for laboratory data collection.
[INDIRECT_PROMPT_INJECTION]: The workflows for colour and viscosity optimization ingest data from external sources (camera images and balance readings) and incorporate an LLM optimization layer. The skill implements strong mitigations against injection risks, including mandatory JSON schema validation, physical constraint checks (e.g., volume sum validation), and requiring explicit user approval before executing any suggestions provided by the optimizer.
[DYNAMIC_EXECUTION]: The skill generates Python-based protocol files for Opentrons robots. These are produced using structured template builders (build_colour_mixing_protocol.py) based on validated numeric inputs. The generation logic is transparent and does not involve executing arbitrary code from untrusted remote sources.

bears-workflows