SAO governs agents as owned, reviewable runtime entities.
The repo still models agents around durable identity material such as:
soul.mdethics.mdorg-map.mdpersonality.md
These documents matter because they define the intended trust and behavioral envelope for an agent, even when the operational control plane is handling identity, sessions, and governed skills.
Each skill should be treated as a capability package with:
- a clear operational purpose
- declared permissions
- reviewable inputs and outputs
- lifecycle status such as pending review, approved, or deprecated
The skills/ directory is kept in the repo to illustrate that governance model.