earth2studio-deterministic-forecast
Warn
Audited by Gen Agent Trust Hub on Jun 12, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The evaluation environment setup script modifies system-wide environment configuration. File: evals/environment/setup/bootstrap.sh. Evidence: echo "PATH=${REPO_ROOT}/.venv/bin:${PATH}" >> /etc/environment. Description: The script appends a path based on the REPO_ROOT variable to the global /etc/environment file. This configuration change applies to all users and processes within the environment. This specific behavior is noted in the skill's BENCHMARK.md as an unsanitized modification of system environment variables.
- [EXTERNAL_DOWNLOADS]: The skill fetches documentation and model weights from external sources. File: SKILL.md. Evidence: References to nvidia.github.io and github.com/NVIDIA for prognostic models and data sources. Description: These downloads are from trusted vendor repositories and are necessary for the skill's primary function of running weather models.
- [PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection. File: SKILL.md. Ingestion points: The agent is instructed to fetch and read content from external URLs to determine current APIs. Capability inventory: The agent uses this information to generate Python scripts using the earth2studio library. Boundary markers: The instructions do not define clear boundaries for the data fetched from these external sites. Sanitization: No sanitization of the fetched documentation content is mentioned.
Audit Metadata