dayan-agent-skills

Human authority ledger

Principle

Before an AI workflow starts, write down which actions remain human-controlled and what evidence is required before each handoff.

Apply it

1. Name every consequential action in the workflow.
2. Mark each action as agent-draft, agent-execute, human-review, or human-only.
3. Attach the evidence required to cross each boundary.
4. Define who can stop, approve, or take over.
5. Keep the ledger with the artifact so later reviewers can audit authority.

Failure pattern

The method has failed when publication, spending, production impact, personal data handling, or external communication happens because no one wrote down where agent authority stops.

Public status

This is a compact public method card, not legal advice and not a substitute for organization-specific approval policy. MIT licensed as part of Dayan Agent Skills.

This site is open source. Improve this page.