design-consultation
Pass
Audited by Gen Agent Trust Hub on May 2, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads the Bun runtime installer from the official bun.sh website to set up required dependencies for visual research tools.
- [REMOTE_CODE_EXECUTION]: Executes the downloaded Bun installer script to configure the local environment. It also utilizes the 'Codex' tool to execute design-related tasks in a remote environment.
- [COMMAND_EXECUTION]: The preamble and workflow utilize shell commands, including 'eval' and 'source', to execute the output of local binaries provided within the 'gstack' environment. These commands manage session state, project slugs, and repository modes.
- [DATA_EXFILTRATION]: Features a 'GBrain Sync' capability that allows the agent to synchronize session memory to a private GitHub repository. It also implements telemetry logging that transmits usage data (skill name, duration, and anonymized device ID) to a remote server. Both functionalities are presented to the user for opt-in consent during the first run.
- [PROMPT_INJECTION]: Includes explicit instructions for the agent to 'Treat the skill file as executable instructions, not reference,' which ensures the model follows the multi-phase design workflow strictly. It also includes directives to ignore generic 'plan mode' behavior in favor of the skill's specific logic.
Audit Metadata