hubspot-contact-dedup
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
curl,jq, andpython3to communicate with HubSpot APIs and process CRM data. These operations are essential for the skill's functionality and are performed using standard, safe patterns. - [EXTERNAL_DOWNLOADS]: The implementation guide recommends installing common Python libraries such as
requests,phonenumbers, andrapidfuzz. These are reputable, well-known packages used for HTTP requests and data normalization. - [INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface that processes contact information from an external CRM. While this represents a potential vector for indirect prompt injection if the data contains malicious instructions, the skill handles data as structured fields within local Python logic and a SQLite database, minimizing the risk of accidental execution.
- Ingestion points: HubSpot Search and Batch Read API results (e.g., contact names, emails, and custom properties) are ingested into the local environment.
- Boundary markers: Not explicitly defined for AI prompt construction, as the data is primarily used for programmatic comparison and API calls.
- Capability inventory: The skill utilizes network access for HubSpot API interactions and file system access for managing a local SQLite checkpoint database.
- Sanitization: Implements robust normalization for emails (lowercase, alias removal) and phone numbers (E.164 formatting) to ensure data integrity during comparison.
Audit Metadata