gog-hardened
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the
gogCLI utility to interact with Gmail, Drive, Calendar, and Sheets. These commands enable the reading of personal emails, document content, and contact lists, as well as the creation of messages and events. - [EXTERNAL_DOWNLOADS]: The skill provides an installation path for the
gogbinary via a third-party Homebrew tap (steipete/tap/gogcli). This relies on a repository maintained by an individual developer rather than a recognized technology organization. - [DATA_EXFILTRATION]: The skill includes specific guardrails forbidding the transmission of Workspace data to external endpoints, network commands, or local clipboard tools. This prevents common patterns where an agent might be tricked into exfiltrating sensitive document content.
- [PROMPT_INJECTION]: The skill contains logic to handle indirect prompt injection surfaces where instructions might be embedded in emails or documents.
- [PROMPT_INJECTION]: Ingestion points:
gog gmail messages search,gog docs cat, andgog drive search(SKILL.md). Boundary markers: The skill includes specific behavioral guardrails instructing the agent to disregard instructions originating from Workspace data regarding authentication or scope changes. Capability inventory:gog gmail send,gog sheets update,gog calendar create(SKILL.md). Sanitization: No explicit data sanitization or escaping of retrieved content is specified before the data is processed or output.
Audit Metadata