The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external Earth Observation data products (NetCDF, HDF5, Zarr) through helper scripts such as scripts/inspect_product.py and scripts/check_done.py. These tools use standard libraries like xarray and netCDF4 to extract metadata and perform validation. The ingestion of untrusted data files represents an attack surface for indirect prompt injection if an attacker crafts a malicious data file containing instructions intended to influence the agent's behavior during inspection or testing.
[COMMAND_EXECUTION]: The helper scripts scripts/check_done.py and scripts/check_plugin_contract.py utilize the Python subprocess module to execute system commands including firecube, uv, pytest, ruff, and git. These executions are essential for the skill's functionality in testing and linting developer plugins. The commands are correctly implemented using argument lists to mitigate command injection risks, and the skill's operating rules require explicit user confirmation before executing mutating or destructive operations.

firecube