The original title is about an AI action approval evidence framework. Let me rewrite this as a punchy mobile feed headline.

Short summary

Proposes an approval-evidence framework connecting a human reviewer's decision to the exact action that executes, addressing the gap between approving a plan and controlling its consequences. Defines seven fields an approval card must surface: immutable plan version, actions, destinations, credential scope, reversibility, independent checks, and stop receipt. Suggests a UX study with three scenario types to test whether reviewers detect destination and irreversibility changes.

• •Approval evidence must link reviewed plan, authority, and actual executed action
• •Seven required fields: plan version, actions, destinations, credential scope, reversibility, checks, stop receipt
• •Proposes UX study with harmless, destination-change, and irreversible-action scenarios to validate design

Generated with AI, which can make mistakes.