month-end-closer
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted financial documents which could contain malicious instructions.
- Ingestion points: Reads invoices, supplier statements, and Trial Balance data via Model Context Protocol (MCP) tools (SKILL.md).
- Boundary markers: Explicitly mandates that untrusted documents be opened with a restricted worker that lacks access to sensitive MCP or write tools.
- Capability inventory: Accesses General Ledger (GL) data and utilizes write tools for schedule generation.
- Sanitization: Implements architectural role separation to isolate the processing of untrusted content from sensitive capabilities.
- [EXTERNAL_DOWNLOADS]: The skill references sub-skills and architectural patterns from the Anthropics financial-services repository on GitHub.
Audit Metadata