airtable
Pass
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes structured data and record comments from Airtable, which are externally controlled and could contain malicious instructions. \n
- Ingestion points: Airtable records, comments, and schema metadata fetched via
records.py,webhooks.py, andschema.py. \n - Boundary markers: Guidelines in
privacy.mdinstruct the agent on data handling, but no technical delimiters are enforced on the ingested content to prevent the agent from obeying instructions embedded in the data. \n - Capability inventory: Extensive CRUD operations (create, update, delete) on records and schemas, and the ability to manage webhooks through scripts. \n
- Sanitization: Relies on natural language instructional guidelines in
privacy.mdfor PII masking and data summarization rather than technical filters.
Audit Metadata