State model
A signed system manifest links code, prompt, model, knowledge, tools, permissions, policies, tests and owner. Any material dependency change creates a new release candidate.
An agent's behaviour depends on model, prompt, tools, knowledge, memory, topology and operating policy, so testing only the model is incomplete.
A certification service versions the complete agent contract, computes its reachable authority and requires risk-tiered evidence before promotion.
Organisations, systems and operating conditions are intentionally anonymised and recomposed. The design demonstrates engineering and banking-domain reasoning; it does not represent a named client estate, vendor product or measured production result.
Typed commands, state changes and receipts cross an event spine without surrendering domain ownership.
Permitted workThe platform supplies reusable controls and contracts. Domain teams retain business logic, data purpose and outcome ownership.
Consistency ruleTie deployed code, prompt, model, data, permissions, policy and evidence to one release manifest and compatibility graph.
Hard boundaryThe model is not a system of record, identity provider, policy authority or proof that an external effect occurred.
Agents, prompts, tools, policies and models must be built, certified, operated, changed and retired as one reachable system.
A changed dependency invalidates affected evidence and can automatically reduce or withdraw authority.
A deterministic outer workflow contains model-led work inside typed, observable calls. Dashed messages remain proposals until policy or a human grants authority.
A signed system manifest links code, prompt, model, knowledge, tools, permissions, policies, tests and owner. Any material dependency change creates a new release candidate.
Build and runtime boundaries use the same schemas. Registry metadata drives discovery, policy, telemetry attribution and deprecation.
The deployed artefact, registry entry and evidence pack share one immutable release identifier; partial promotion is not a valid state.
These roles are deliberately vendor-neutral. Each can be independently owned, versioned and replaced.
Returns candidate entities and typed relationships with match features, contradictions, effective dates and non-merge evidence.
Evaluates identity, purpose, capability, amount, risk tier and policy version; returns allow, deny, step-up or human-review with reasons.
Appends request, versions, policy result, model proposal, approval, action receipt, readback, correction and custody events under one correlation key.
Runs component, route, trajectory, failure, harm and outcome tests against the versioned system manifest.
Durable records carry provenance, authority, effect and custody without turning a transcript into an uncontrolled memory store.
Tie deployed code, prompt, model, data, permissions, policy and evidence to one release manifest and compatibility graph.
The selected design is not universally superior. It is the safer fit for this boundary and failure cost.
Bias consequential journeys against false merge and retain unresolved candidates.
Automatically merge the highest-scoring candidate.
Cost acceptedMore cases require clarification, but one person's authority or risk cannot silently attach to another.
Compile stable decision logic and retain retrieval for explanation and residual ambiguity.
Ask a model to interpret the source document for every request.
Cost acceptedRule compilation needs controlled change, but creates repeatable decisions, regression tests and clear exceptions.
Use append-only events plus a rebuildable current-state projection.
Overwrite the case row with its latest status.
Cost acceptedReplay and storage are more complex, but point-in-time reconstruction and correction lineage remain possible.
Test prompts, models, tools, knowledge, policies, state transitions and human paths together.
Use a static answer-quality benchmark as the release gate.
Cost acceptedSystem evaluation takes longer and needs synthetic environments, but detects authority and recovery failures that answer scoring misses.
Retries are bounded by knowledge of business effect; unknown outcome remains visible, owned and independently reconciled.
Actual thresholds belong to accountable service owners. The design exposes the equations and observables that those owners must baseline.
gate_runtime = test_cases x routes x dependency_versionsrelease_capacity = available_environments / average_gate_durationoperating_cost = model + tools + platform + human_review + incidentsEnforce purpose, tenant and workload identity at every hop; telemetry stores metadata rather than unrestricted prompts or payloads.
A design is production-ready only when teams can prove what happened, recover it and change it safely.
reachable-authority analysis
cross-version compatibility
failure and rollback rehearsal
cost, latency and trace attribution
Agent manifest, dependency graph, capability tokens, evaluation pack, approvals, release receipt and monitoring baseline.
Product, risk and service owners certify, constrain or retire the agent.
Automate evidence collection before automating approval. Expand self-service only when the paved route proves safer and faster than bespoke delivery.
Platform, domain product, data, security, model-risk and service owners share certification decisions.