human-approval-boundary

What problem does it solve?

This Skill prevents high-risk actions from being carried out on assumed consent. It forces a stop before destructive, security-sensitive, or irreversible work crosses a boundary, so the human can approve with full context.

Core Features & Use Cases

• Pre-action halting: Stops before schema changes, destructive migrations, production data changes, secret handling, deployments, releases, billing impacts, git history rewrites, or broad refactors.
• Stop-when-unclear safety: Halts when the security or data impact cannot be stated confidently, or when ambiguous instructions could lead to materially different outcomes.
• Structured approval requests: Produces a concise request with the exact action, boundary type, blast radius, reversibility, options, and a recommendation.
• Use case: A migration is about to drop a table, or a change might affect session token validation, and the Skill interrupts execution until explicit human approval is received.

Quick Start

Ask the human for explicit approval before proceeding with the risky action, including the exact change, blast radius, reversibility, and available options.