geo-measure
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external engine observation bundles (bundle.json) to generate reports. This ingestion of untrusted data creates a surface for indirect prompt injection if the input files contain instructions designed to manipulate the agent's output.
- Ingestion points: bundle.json referenced in SKILL.md and scripts/run_measure.py.
- Boundary markers: The workflow specifies using 'approved, file-backed engine observation bundles' and explicit rejection criteria in references/measurement-method.md.
- Capability inventory: Execution of scripts/run_measure.py, file system read access for inputs, and write access for reports in the runs-root directory.
- Sanitization: The wrapper script does not implement additional sanitization of the JSON input.
- [COMMAND_EXECUTION]: The skill's primary workflow involves executing a Python script (scripts/run_measure.py) to process data. This is an expected part of the skill's functionality for offline measurement.
Audit Metadata