stakeholder-map

Pass

Audited by Gen Agent Trust Hub on Nov 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources such as Slack messages, call transcripts, and email exchanges to identify stakeholders and assess their stance.
  • Ingestion points: Data is gathered via Salesforce GraphQL queries, Slack tool outputs, email tool outputs, and document tool outputs.
  • Boundary markers: The skill instructs the agent to ground all data and "cite every value exactly as queried," which helps ensure the agent relies on retrieved data rather than fabricated or injected instructions.
  • Capability inventory: The skill uses dispatch_readonly for data retrieval, dispatch for record creation/updates, and display_widget for visualization.
  • Sanitization: The skill emphasizes using grounded human labels and exact record links, though it does not explicitly define sanitization for free-text content in transcripts or messages.
  • [SAFE]: The use of a GraphQL query to fetch the currentUser Id as a connectivity check is a standard and safe diagnostic tool to verify the session's authentication status.
  • [SAFE]: The instructions to execute tool calls silently and terminate output after a successful widget display are platform-specific UI/UX constraints designed to optimize the end-user experience.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 11, 2026, 06:39 PM