The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external configuration files and measurement data, which constitutes a potential attack surface for indirect instructions.
Ingestion points: The script scripts/simulate_battery.py loads protocol definitions from JSON files and measurement data from user-supplied CSV files.
Boundary markers: Input data is parsed using standard libraries and strictly validated against a defined schema. The validate_protocol function ensures that only allowed fields are processed.
Capability inventory: The skill's capabilities are limited to local numerical computations and writing results to CSV/JSON files in a user-defined output directory. It does not possess capabilities for arbitrary command execution, network exfiltration, or persistence.
Sanitization: The script enforces strict type-checking for parameter overrides, requiring them to be finite numeric values. This prevents the injection of executable logic or functions through configuration fields.
[COMMAND_EXECUTION]: The script performs local numerical simulations by invoking the PyBaMM scientific computing library.
Execution method: The scripts/simulate_battery.py script manages local simulation runs and numerical sensitivity analyses via Python library calls.
Security controls: Telemetry is explicitly disabled at the start of the script via the PYBAMM_DISABLE_TELEMETRY environment variable to ensure data privacy during execution.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions for installing the necessary scientific software environment.
The skill references standard electrochemical modeling packages (pybamm and pybammsolvers) for installation via the uv package manager. These are well-known tools in the battery research community.

pybamm