migration-parity-audit
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions contain strong safety guardrails for migration auditing. It explicitly commands the agent NOT to call production systems, NOT to copy customer data, and NOT to perform destructive actions like stopping services or switching traffic during an audit. The logic focuses on local synthetic fixtures and behavioral verification without evidence of prompt injection, data exfiltration, or unauthorized command execution. All evaluated evaluation prompts (
evals/) correctly test for these safety boundaries by requiring the agent to refuse dangerous or premature activation commands.
Audit Metadata