Docs

Category 05 of 12

Reliable by Design

Inside the self-healing workflow engine: the recovery ladder, the SWIFT verdict that separates completed from accomplished, and the node reference.

Pages in this category

Completion Is Not Accomplishment

The SWIFT verdict engine probes the real world after a run settles instead of trusting a self-report, distinguishing accomplished from merely completed.

SWIFTVerdict
The Recovery Ladder: Five Rungs From Reflex to Escalation

How a failed node recovers: deterministic rebind, capability-aware remedy, config-aware surfacing, certified structural repair, then honest escalation.

Self-healRecovery
Blueprint Law: Blessing, Restructuring & Rollback

A proven, blessed workflow graph is never autonomously restructured — repair requires certification and a rollback path.

BlueprintGovernance
The Execution Model: Ticks, Ready Queue & Waiting Inputs

Inside WorkflowEngine: the ReadyQueue, the WaitingInputBuffer, and what happens on every tick from dispatch to snapshot.

EngineTicks
Objectives & Acceptance Checks

How a run-scoped Objective declares the acceptance checks that define done, verified against real outputs rather than a self-report.

Objectives
Cognitive Looping: Converge & Pursue

Converge iterates a cohort sub-graph until a continuation policy stops it; Pursue is converge with ASSESS and REFLECT turned on.

ConvergePursue
ASSESS & REFLECT: How the Loop Grades and Corrects Itself

A graded 0..1 progress signal for stall detection, and a bounded self-critique pivot ladder for when a loop stalls.

ASSESSREFLECT
Workflow Design Doctrine: The Iron Rules

The doctrine injected into both the synthesis and reviewer prompts, teaching agents to design for failure and control flow, not just structure.

Design doctrine
The Pattern Library & Self-Improving Playbook

Static named failure/control-flow patterns, plus a living, workspace-scoped playbook of failure-mode-to-fix lessons recalled at build time.

PatternsPlaybook
Robustness Audit & Preflight Validation

The deterministic enforcement layer that audits an authored graph for robustness before it ever runs.

RobustnessPreflight
The Instinct Engine

Detects repeated workflow failure patterns, writes the lesson into the Brain, and emits an operator-visible proposal.

Instinct Engine
Atomic Evolution: Changing a Live Run Safely

Mid-run graph changes must pass the same contracts as authoring — no data-coupling break, no gutted capability, never a regression below the graph's proven green.

Atomic EvolutionAgent-primary
The Eval Harness: Grading Outcomes, Not Self-Reports

The regression gate that runs cold-start build tasks against a seeded workspace and grades the actual outcome — pass@k versus pass^k.

Eval harnessAX
Capability Router: One Door for Workflow, Node, Phase, Agent & MCP

Resolves a single capability URN onto one execution path, so chat, workflow, and MCP all delegate through the same trusted handlers.

CapabilityRouter
Triggers, Cron & the Listener Runtime

Manual, cron (natural-language to UTC), webhook, and persistent-listener triggers — how a run actually starts.

TriggersListeners
Replay & Partial Recovery

Restarting a run from a node, a failed branch, an edited node, or a checkpoint — deterministic partial replay.

Replay
Worktree Isolation for Swarms & Converge

How parallel agent work gets its own git worktree, so a swarm or a converge cohort never steps on another branch's changes.

WorktreesSwarm
Node Reference: Control Flow

router, merge, parallel, wait, checkpoint, phase_gate, subflow, loop, stop_error, return_output — every control-flow node explained.

NodesReference
Node Reference: Intelligence Nodes

agent_task, agent_session, agent_swarm, converge, pursue, planner, evaluator, guardrails — every intelligence node explained.

NodesReference
Node Reference: Data, Compute & IO

transform, code, data_query, http_request, aggregate_window, knowledge, browser, artifact_save, mcp, human_input — every data/compute/IO node explained.

NodesReference

Continue