Pipeline Focus Task Planner
Pass
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: A thorough analysis of the skill's instructions and methodology documents identified no malicious patterns or security risks.
- [PROMPT_INJECTION]: The skill processes untrusted deal data from CRM systems (pipeline_deals), creating an indirect prompt injection surface. However, this is a standard functional requirement for CRM integration and no exploitative instructions were found. Evidence: Ingestion points: 'pipeline_deals' input; Boundary markers: None; Capability inventory: 'crm'; Sanitization: None.
- [DATA_EXFILTRATION]: Data access is limited to the CRM capability required for task planning. No patterns of unauthorized data exfiltration, network requests to non-whitelisted domains, or access to sensitive local environment files were detected.
- [REMOTE_CODE_EXECUTION]: The skill logic is entirely instructional and does not download or execute external code or scripts.
Audit Metadata