Granola Meeting Notes
Pass
Audited by Gen Agent Trust Hub on Jul 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local script
granola.jsincluded in the plugin root to list and display meeting data. - [DATA_EXFILTRATION]: The skill accesses sensitive personal data by reading meeting notes and transcripts from the Granola app's local macOS cache.
- [PROMPT_INJECTION]: The skill processes untrusted data from meeting transcripts, which can lead to indirect prompt injection.
- Ingestion points: Data enters the context via
list,show,search, andexportcommands inSKILL.md. - Boundary markers: None provided in the instructions.
- Capability inventory: Execution of subprocesses via the
granola.jstool. - Sanitization: No evidence of data sanitization before processing.
Audit Metadata