weekly-review
Pass
Audited by Gen Agent Trust Hub on Jul 22, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to process untrusted data from external sources (emails and calendar invites). It effectively mitigates indirect prompt injection risks through a set of mandatory 'Safety rules (HARD)'.
- Ingestion points: Gmail and Google/Microsoft 365 calendar connectors (SKILL.md).
- Boundary markers: The agent is instructed to summarize content and present findings as 'proposed' or 'inferred' rather than absolute facts, ensuring the user reviews the context (SKILL.md, review-framework.md).
- Capability inventory: Read access to Gmail and Calendar; write access to Notion (SKILL.md).
- Sanitization: The skill contains explicit instructions to treat all email/invite content as hostile, ignore any instructions found within them, and never auto-fetch embedded links or images (SKILL.md).
- [DATA_EXFILTRATION]: The skill follows the principle of least privilege. It uses read-only connectors for its primary data sources and requires explicit user confirmation before writing any summary to Notion. It specifically prohibits opening or fetching external URLs found in content, which prevents common automated exfiltration techniques.
- [COMMAND_EXECUTION]: No shell commands, dynamic script execution, or '!' command patterns were found in the skill or its references.
Audit Metadata