callable-lead-count
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external data (contact names, job titles, and company information) retrieved from HubSpot to classify leads and generate reports. This represents a potential surface for indirect prompt injection where malicious instructions embedded in CRM data could attempt to influence the agent's behavior.
- Ingestion points: Data retrieved from HubSpot via the
search_crm_objectstool (Stage 1). - Boundary markers: The skill does not explicitly define delimiters to isolate external contact data from its internal reasoning instructions.
- Capability inventory: The skill utilizes the
search_crm_objectstool for data retrieval and performs local file writes to~/.claude/skill-analytics/for telemetry (Stage 8). - Sanitization: While the skill uses keyword-based logic to categorize job titles (Stage 3), there is no evidence of broader sanitization or escaping of contact fields before they are interpolated into the final markdown report or the JSON integration payload.
Audit Metadata