estate-and-tax-brief
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Sensitive Financial Data Handling: The skill aggregates and processes high-value information, including tax return actuals (AGI, taxable income, and effective rates), estate document status, and account titling/beneficiary designations. This data is essential for the skill's primary function of identifying mismatches between estate plans and actual account setups. The skill manages this sensitivity by restricting operations to connected financial platforms and using specialized subagents for data extraction.
- Indirect Prompt Injection Surface: The skill ingests data from external sources that could theoretically contain instructions intended to influence the agent's behavior, such as CRM notes or manually uploaded document summaries.
- Ingestion points: The skill reads household CRM notes (Wealthbox/Redtail/Salesforce) and accepts manual uploads or text pastes of estate and tax summaries as fallbacks for missing connectors (SKILL.md, Steps 1, 2, 3, and 4).
- Boundary markers: The skill utilizes specialized subagents for data extraction and comparison tasks, and it explicitly mandates the inclusion of "not-legal-advice" disclaimers in all generated outputs to maintain clear contextual boundaries.
- Capability inventory: The skill has the capability to generate formatted files (Markdown, DOCX, PDF) and create follow-up tasks within a connected CRM system (SKILL.md, Steps 7 and 8).
- Sanitization: All state-changing actions (specifically CRM task creation) are gated by mandatory advisor approval. The skill also follows a "Cite, never compute" rule to ensure data integrity and prevent the fabrication of financial figures.
Audit Metadata