Deal Rescue Plan

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by processing unstructured data from external CRM records.
  • Ingestion points: The skill ingests untrusted content through the deal_context input and the get_deal and get_deal_activities tool calls, which retrieve meeting notes and email history from the CRM.
  • Boundary markers: There are no explicit delimiters or specific instructions in the prompt logic to differentiate between system instructions and the ingested external data.
  • Capability inventory: The skill is granted the crm capability, allowing it to read deal records and activities and suggest new tasks based on the analysis.
  • Sanitization: The skill does not perform explicit sanitization, filtering, or validation of the retrieved CRM content prior to processing it for diagnosis.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 11:37 AM