accrual-schedule
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface by processing external, untrusted data from invoices and vendor statements.
- Ingestion points: External invoices and vendor statements are processed to extract amounts used in calculations (SKILL.md).
- Boundary markers: The prompt includes an explicit warning that these sources are untrusted and should only be used for data extraction.
- Capability inventory: The skill utilizes an internal General Ledger (GL) tool (internal-gl MCP) for read access and generates text-based journal entry drafts. It lacks the capability to post transactions, write files, or perform arbitrary network operations.
- Sanitization: The skill logic relies on a separate extraction step and focuses on applying internal policies. The output is explicitly labeled as a staged draft requiring controller sign-off, providing a human-in-the-loop security control.
Audit Metadata