meta-agent-sla-financial-controls
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is purely informational and procedural. It outlines workflows, approval matrices, and policy frameworks for financial auditing and variable-revenue management.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external business artifacts such as SLA contracts, signed agreements, and ERP records (referenced in
SKILL.mdunder 'Required Inputs'). - Ingestion points: Processes contract terms, approval roles, and financial records provided at runtime.
- Boundary markers: The skill includes explicit instructions to 'label assumptions', 'cite sources', and 'remain read-only until explicit authority is granted'.
- Capability inventory: Capabilities are limited to non-mutating checks and drafting narratives. No file-write or network-send tools are requested or used.
- Sanitization: Relies on human-in-the-loop review by a 'named decision-maker' and 'Controller' for all outputs.
Audit Metadata