crm-automation
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The workflows described in the skill ingest untrusted data from external webhooks, such as website lead forms and product analytics events. This data is directly interpolated into prompts sent to the OpenAI API (Workflow 1) to classify leads and suggest communication topics.
- Ingestion points: External data entering via 'Webhooks > Custom webhook' modules described in SKILL.md.
- Boundary markers: The instructions use standard chat completion roles but lack explicit boundary markers or 'ignore embedded instructions' directives for the dynamic lead properties.
- Capability inventory: The skill describes capabilities to create records in Pipedrive, send emails via SendGrid, and post to Slack (Workflow 1 and 2).
- Sanitization: There is no mention of sanitizing or escaping the external content before interpolation.
- [DYNAMIC_EXECUTION]: Workflow 2 includes a JavaScript code block designed for an N8N 'Code' node. This script processes runtime data to calculate churn risk levels. While this is standard functionality for the platform, it involves dynamic logic execution based on incoming external data.
Audit Metadata