explore-code

Warn

Audited by Gen Agent Trust Hub on May 29, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The file scripts/write_outputs.py performs dynamic code loading and execution of a module located at a relative path outside the skill's directory.
  • Evidence: The script calculates a path to ../../../shared/scripts/write_explore_bundle.py using Path(__file__).resolve().parents[3] and executes it using spec.loader.exec_module(module).
  • Context: While this is a common pattern for shared utilities in certain repository structures, it introduces a dependency on the host environment's file system integrity and is classified as dynamic execution from a computed path.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 29, 2026, 07:59 PM
Security Audit — agent-trust-hub — explore-code