Governance

Static governance for identity, rules, skills, and guidance.

This page defines the static governance layer: who the agent is, what it can use, what rules it follows, and how those choices are released.

identity
keys
scoped rules
skills
guidance
releases
Static layer
Stable enough to trust, small enough to explain.

Know exactly what is governed before mission control and task execution are introduced.

Identity

Treat identity as trust, not just a label.

Identity covers practical trust surfaces: enrollment, keys, ownership, and which client is speaking.

Agent identity

Stable identifiers for each agent and the workspace it belongs to.

Keys

SSH, API keys, and other enrollment material needed to trust the client.

Ownership

Who enrolled the agent and who can rotate or revoke access.

Status

What the control plane expects versus what the client reports.

Rules

Scoped rules make inheritance obvious.

Rules are the hard or soft behavior layer: scannable, scoped, and explainable.

Hard rules

Non-negotiable boundaries with clear enforcement.

Soft guidance

Useful defaults and preferences that can be overridden.

Scope chain

Global, team, project, workspace, agent, with visible overrides.

Release history

Every change has a version and a reason.

Skills

Skills stay separate from rules and guidance.

Skills are folder-based work units. Keep them portable, versioned, and easy to inspect.

Folder structure

SKILL.md, scripts, references, and assets in one package.

Private skills hub

A team can maintain internal skill sets without turning them into policy.

Release control

Pin, enable, disable, or roll back a skill version.

Compatibility

The same artifact can be mapped to different runtimes when supported.

Cluster pages

Downstream pages for trust, policy, skill packaging, and guidance.

Move from governance into focused pages for each managed object.