The Agent Skills Directory

[SAFE]: The skill is designed for per-tick signal computation in a weather-trading pipeline. Its operations are strictly limited to reading three local JSON state files and writing one output file using the internal weather_runtime module.
[SAFE]: No network calls, credential handling, or external code execution patterns were identified. The skill explicitly limits its I/O to defined paths within the runtime/state/ directory.
[SAFE]: The skill processes data ingested from previous pipeline steps (forecasts and market data). Although it lacks explicit prompt boundary markers for this data, its capabilities are restricted to mathematical calculations and atomic file writes, which prevents the data from being used to trigger unauthorized actions.
[SAFE]: Robust failure handling is implemented, ensuring the agent aborts execution if input data is malformed rather than proceeding with potentially corrupted state.

signal-gen