skills/mverab/egeoagents/geo-loop/Gen Agent Trust Hub

geo-loop

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute specific CLI tools (egeo loop run, egeo loop doctor, egeo loop collect) and a linting script (python -m egeo.substrate_lint). These are specialized utility commands for the eGEOagents ecosystem used to manage the workspace state and validate file structures.
  • [DATA_EXPOSURE]: The skill requires a BRAVE_API_KEY for search result collection. It correctly instructs the agent to retrieve models and budget configurations from a local config.yaml rather than hardcoding credentials or prompting for them insecurely.
  • [SAFE]: The skill enforces a strict boundary policy (Non-negotiable #1), explicitly prohibiting writes to the repository tree and directing all output to a user-defined workspace directory ($EGEO_HOME, default ~/.egeo/). This prevents unintended modification of the agent's own code or logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 04:35 PM
Security Audit — agent-trust-hub — geo-loop