atmospheric-science-research-skill

Pass

Audited by Gen Agent Trust Hub on Mar 13, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is dedicated to scientific research tasks using standard Python data science libraries.
  • [COMMAND_EXECUTION]: No unauthorized command execution patterns or shell access vectors were detected. Code standards focus exclusively on scientific computation.
  • [DATA_EXFILTRATION]: No network activity or unauthorized data access patterns were identified; the skill operates on local datasets for research purposes.
  • [REMOTE_CODE_EXECUTION]: The skill does not download or execute remote scripts or binaries.
  • [PROMPT_INJECTION]: No attempts to bypass safety filters or override system instructions were found.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes scientific data files (NetCDF, GRIB) using established libraries like xarray and does not interpolate untrusted data into prompts or sensitive command contexts.
  • Ingestion points: Binary data ingestion is defined in references/compute-standards.md via xr.open_dataset.
  • Boundary markers: None (typical for structured scientific formats).
  • Capability inventory: Limited to data transformation and image generation.
  • Sanitization: Standard library usage for data parsing provides inherent protection against non-executable data formats.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 13, 2026, 02:19 PM