HC10 · Human Control & Oversight

Network Degraded State

Purpose Shows a multi-agent run in partial failure — which agents completed, which stopped, which are still running, and whether the orchestrator considers the run recoverable. Exposes the blast radius of a partial failure before the human decides how to respond.

Use when One or more agents in a multi-agent network has failed, errored, or stopped while others are still running or have completed. The run is in a degraded but not necessarily terminal state. Also use when an agent's failure blocks downstream agents from proceeding.

Do not use For single-agent errors — use HC06 Recovery & Override instead. Do not use when all agents have failed — that is a total run failure, not a degraded state. Do not collapse the degraded state into a generic error message.

Failure mode prevented Partial failures being invisible — runs continuing in a broken state because the human was never shown which agents failed and what the downstream consequences were.

Component UI

Component UI coming soon

Overview

Design rule

A partial failure is a governance event, not just a technical error. The human must see the full failure footprint — what failed, what's still running, what's now blocked — before deciding whether to continue, repair, or abort.

The Network Degraded State component surfaces when a multi-agent run has entered partial failure. It is not a generic error screen — it is a structured representation of which agents are in which state, how those states affect each other, and what the orchestrator has determined about recoverability. The human sees a complete picture of the network, not just the failing node.

Critically, this component must distinguish between: agents that have completed successfully, agents that are still running, agents that have failed or stopped, and agents that are blocked because a dependency failed. These are not all the same state and must not be collapsed into a single "failed" indicator.

Anatomy

Network status header: Overall run state (Degraded / Partially Complete / Blocked) with timestamp
Agent status list: Each agent with its current state — Completed, Running, Failed, Blocked
Failure detail: For failed agents, the error type and what step it stopped at
Dependency impact: Which running or pending agents are blocked by the failure
Recoverability indicator: Orchestrator's assessment — Recoverable / Unrecoverable / Unknown
Recovery controls: Links to HC06 Recovery & Override for next actions

Accessibility

Related Components

HC03 Agent State Indicators — shows per-agent state during normal execution
HC06 Recovery & Override — provides recovery actions after degraded state is surfaced
HC05 Agent Attention Triggers — may trigger the degraded state display when anomaly is detected
HC11 Conflict Resolution Surface — handles conflicting outputs, which may co-occur with degraded state
Pattern 03: Multi-Agent Orchestration — Stage 03: Parallel Execution State
Pattern 04: Multi-Agent Governance — Stage 01: Partial Failure Isolation

← Previous Spawn-Time Consent Gate Next → Conflict Resolution Surface