safe-write-operations
Pass
Audited by Gen Agent Trust Hub on May 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill acts as a security policy, enforcing a mandatory four-step protocol (State, Blast Radius, Confirm, Execute) for any operation marked with 'requires_approval: true'. This ensures that the agent cannot autonomously execute destructive or high-cost actions without direct user oversight.
- [SAFE]: Implements defensive constraints to minimize the blast radius of errors, such as a 50% limit on budget modifications in a single call and a requirement to re-confirm batches larger than five operations.
- [SAFE]: Authentication error handling correctly directs users to the official vendor domain (markifact.com) for connection management, which is appropriate and consistent with the skill author's context.
Audit Metadata