Policies

Policies are declarative rules that govern what agents and humans can do within a matic org. Each policy defines a scope, trigger conditions, and an enforcement mode of block, warn, or log, and is stored as a markdown file with YAML frontmatter in .matic/policies/. The Daemon evaluates policies during signal interpretation, activation resolution, and context assembly, while the Runtime enforces them through guardrail hooks at execution time.

Scope & Inheritance

Scope & Inheritance explains how policies cascade through the org -> project -> team -> agent hierarchy, and the rule that lower scopes can tighten but never loosen a parent policy.

Trigger Conditions

Trigger Conditions covers the condition language policies use to match actions, including operators like IS, MATCHES, EXCEEDS, and WITHIN applied to fields such as action, branch, budget_consumed, and time_of_day.

Policy Lifecycle

Policy Lifecycle describes how a policy moves from draft to active to retired through authoring, review, activation, and versioning, and why every update produces a Decision record.

Policy Proposals

Policy Proposals shows how agents propose new policies as Proposal primitives in response to signals, and how those proposals move through human review before taking effect.

Notification Policy

Notification Policy defines the event-to-recipient-to-channel mappings that control who gets notified, which channels are used, and how escalation chains progress when acknowledgements are missing.

Sandbox Policy

Sandbox Policy details the isolation rules for sandbox mode, including suppressed side effects, separate budget limits, no production channel interaction, and the graduation path to live execution.

Violation Consequences

Violation Consequences explains what happens when a policy is violated, including enforcement modes, HITL escalation, agent suspension, override permissions, and the audit trail each violation produces.