← Agent catalogue
Terminus (architecture validation)

Preview portrait

Integration

Terminus (architecture validation)

Read-only governance signal for operators and boards; complements roadmap vision nodes.

LiveCatalogue status
ObservingRuntime posture

Deterministic validation of declared Agent Overview topology and catalog against Prisma `ExecutionJobStage` and leadership-chain rules — explains drift; does not auto-fix.

  • Architecture governance
  • Terminus
  • Validation
  • Non-mutating

Expanded capabilities

  • Assess architecture governance and optimisation posture
  • Complement TERMINUS substrate observation without dispatching work
  • Surface optimisation candidates as RECORD_ONLY / reviewable items
  • Keep recommendations human-applied
  • Stay honest about in-build vs live readiness

How they learn & improve

The validator improves recommendation quality from operator accept/reject of optimisation candidates — it never force-promotes architecture changes.

  1. Observe declared vs operational architecture signals.
  2. Propose optimisation candidates for human review.
  3. Apply only through explicit operator/product change paths.

How they work together

Terminus Validator partners with TERMINUS (substrate) and TOPOLOGY (graphs) to recommend — humans still change production architecture.

Practical benefits

  • Optimise without autopilotYou get ranked architecture suggestions that still need a human to land.
  • Clear split of dutiesSubstrate health (TERMINUS) vs architecture advice (validator) don’t blur into one false green.
  • Reject bad adviceDeclined candidates teach the next review cycle without changing prod on their own.

Learning and improvement stay human-visible: candidates are RECORD_ONLY until an operator reviews and applies them. Sensitive sends, publishes, and approvals never run silently.

Terminus (architecture validation) — AgentIQ Agent Catalogue · AgentIQ Labs