canary
Pass
Audited by Gen Agent Trust Hub on Apr 1, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill's setup routine includes a command to install the Bun runtime by piping an installation script from the official bun.sh domain directly to the shell.
- [EXTERNAL_DOWNLOADS]: Connects to user-defined application URLs for health monitoring and retrieves documentation from external sources like garryslist.org during initial setup.
- [COMMAND_EXECUTION]: Invokes several local binaries and shell scripts within the gstack skill directory (e.g., gstack-config, gstack-telemetry-log) to manage session state and user preferences.
- [DATA_EXFILTRATION]: Usage metrics, including skill execution results and the name of the active repository, are logged to a local analytics directory for telemetry purposes.
- [PROMPT_INJECTION]: The skill scrapes and parses untrusted navigation links and page text from the monitoring target URLs. This creates an indirect prompt injection surface where external content could influence the agent's monitoring logic, although the risk is limited by the tool's observation-focused capabilities and boundary markers in the report generation process.
Audit Metadata