planetscale-autonomous-execution-mode
Pass
Audited by Gen Agent Trust Hub on Jul 5, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill establishes a rigorous safety protocol for autonomous execution, requiring an explicit 'activation contract' from the operator that includes risk acknowledgment, defined scope, and environment verification.
- [SAFE]: Implements a mandatory refusal policy for 'Class E' operations (e.g., dropping production databases, exposing secrets, or removing safety mechanisms), ensuring these high-risk actions are never executed autonomously regardless of user phrasing.
- [SAFE]: Defines comprehensive halt conditions that trigger during failures, verification errors, or detected state drift (external changes occurring during the run), prioritizing database integrity over completion.
- [SAFE]: Mandates a 'verify-after-each-step' approach where the agent must read back the live database state to confirm a mutation's success before proceeding to the next atomic task.
- [SAFE]: Provides a structured framework for 'Standing Authorizations' for scheduled tasks, requiring committed, reviewable artifacts with expiration dates and bounded operation allowlists.
Audit Metadata