zcode-subagent

Pass

Audited by Gen Agent Trust Hub on Jul 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The scripts/zcode_agent.py script utilizes the subprocess module to execute system commands necessary for orchestrating the ZCode application and its environment. This includes querying the global NPM root to locate dependencies, using macOS open and osascript utilities for application management, and running the agent-browser CLI to interact with the Electron-based UI.
  • [EXTERNAL_DOWNLOADS]: The skill requires external software to function, specifically the agent-browser package from the NPM registry and the ZCode desktop application from the official vendor domain (zcode.z.ai). These downloads are documented as prerequisites and are standard for the tool's automation workflows.
  • [DATA_EXFILTRATION]: The skill accesses application-specific data stored in ~/.zcode/cli/db/db.sqlite and local JSONL log files. This access is used to extract session history and tool telemetry to provide a concise digest of the subagent's performance back to the calling agent. No evidence of unauthorized data transmission to external domains was found.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 15, 2026, 06:06 PM
Security Audit — agent-trust-hub — zcode-subagent