entity-memory
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill integrates with several well-known third-party services and requires their respective client libraries, including @hubspot/api-client, jsforce, pg, and mysql2.
- [EXTERNAL_DOWNLOADS]: The skill uses the vendor's official SDK, @personize/sdk, for interaction with the Personize API.
- [DATA_EXFILTRATION]: As part of its core functionality, the skill transmits entity-related data (e.g., contacts and company records) to the vendor's API at api.personize.ai.
- [PROMPT_INJECTION]: The skill processes data from external sources such as CRMs and databases, creating a surface for indirect prompt injection where instructions could be embedded in processed records. Ingestion points include the 'content' and 'rows' fields in the memorize actions.
- [PROMPT_INJECTION]: The documentation for the memorize action describes a pattern called 'Extraction Hints,' which involves prepending natural language instructions to content to steer AI extraction behavior.
- [SAFE]: The skill follows security best practices by handling sensitive credentials like API keys and database strings through environment variables in all provided recipes and deployment configurations.
- [SAFE]: The Dockerfile and CI/CD templates include explicit security warnings and practices to prevent the inclusion of sensitive environment files within built container images.
Audit Metadata