airtable
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMNO_CODEPROMPT_INJECTION
Full Analysis
- [NO_CODE] (INFO): The skill does not include any executable scripts (Python, JS, or Shell); it consists of documentation and metadata that directs execution to a remote Refly workflow.\n- [Indirect Prompt Injection] (MEDIUM): The skill defines a significant vulnerability surface for indirect prompt injection.\n
- Ingestion points: The skill ingests untrusted data via the
record_dataandquery_filterparameters inSKILL.md, which are used to interact with external Airtable databases.\n - Boundary markers: No delimiters or instructions to ignore embedded commands are present in the prompt templates.\n
- Capability inventory: The skill can create and query database records, which constitutes a write/exfiltration capability when combined with untrusted external content.\n
- Sanitization: There is no evidence of input validation, schema enforcement, or output escaping for data retrieved from or sent to the external database.
Audit Metadata