Work Model

The work model defines how matic turns inbound events into executable work, tracks that work through completion, and records every execution for audit and replay. It covers the full lifecycle: from a signal arriving at the system boundary, through activation and task decomposition, to agent-executed runs that produce verified artifacts. Understanding this model is essential to operating matic because it determines how agents receive work, what context they see, and how delivery is validated.

Signals

Signals define the atomic input unit emitted by channels, routines, webhooks, and system events, carrying a normalized payload of text, data, files, and references into the interpretation pipeline.

Activations

Activations describe the resolved intent-to-work: when the signal pipeline determines that a signal warrants action, it produces an activation, a decision-recorded commitment that transitions the system from interpretation into planning.

Tasks

Tasks cover the semantic specification of discrete work. They answer what needs to happen, why, and to what standard, but are not directly executable and decompose into work items during planning.

Work Items

Work Items are the atomic, agent-executable units. Each carries exact scope, a responsible agent, acceptance criteria, a floor/ceiling work contract, and time estimates, then lives in an agent's work pile until executed.

Runs

Runs are the immutable execution records for work item execution, capturing the context envelope snapshot, agent output, terminal status, and delivery flags needed for inspection, comparison, and replay.

Context Envelope

Context Envelope explains the complete, ordered input assembled for an agent at execution time, including persona, active skills, work item spec, relevant memory, applicable policies, runtime state, and, when signal-triggered, the originating signal payload.

Staffing and Dispatch

Staffing and Dispatch explains how agents are matched to work and how parallel execution is coordinated. Staffing composes teams by evaluating agent probes against project requirements, while dispatch schedules independent work items for concurrent execution within budget constraints.