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.
Teams often repeat security, risk and operational evidence manually at each environment, creating inconsistent decisions and slow promotion without improving control quality.
A delivery service compiles reusable evidence from the versioned agent contract, adds environment-specific tests and routes only genuine exceptions to accountable reviewers.
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.
Five operating planes separate interaction, identity, decision control, authoritative state and operating evidence.
Permitted workThe delivery platform automates evidence and enforces gates. Named owners approve risk, exceptions and production authority.
Consistency rulePromote signed, immutable manifests and keep runtime registry, deployed artefact and evidence pack on the same version.
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.
Evidence reuse requires unchanged dependencies, compatible environment controls and a valid approval purpose; expired evidence cannot be silently carried forward.
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.
Evaluates identity, purpose, capability, amount, risk tier and policy version; returns allow, deny, step-up or human-review with reasons.
Runs a versioned state machine with explicit waits, deadlines, retries, compensations, human tasks and terminal states.
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.
Promote signed, immutable manifests and keep runtime registry, deployed artefact and evidence pack on the same version.
The selected design is not universally superior. It is the safer fit for this boundary and failure cost.
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.
Keep consequential state transitions deterministic and use models inside bounded steps.
Let the model choose the complete path and recovery sequence.
Cost acceptedThe shell reduces flexibility, but makes deadlines, retries, permissions and recovery testable.
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 + incidentsKeep production data and secrets out of build evidence; use synthetic, masked or approved test sources with retained provenance.
A design is production-ready only when teams can prove what happened, recover it and change it safely.
scope and owner completeness
functional, safety and adversarial suites
NFR and recovery evidence
canary, rollback and retirement proof
Build manifest, dependency fingerprint, inherited evidence, environment deltas, exception decisions, approvals and promotion receipt.
Named control owners decide exceptions; automation only assembles and verifies the evidence chain.
Automate evidence collection before automating approval. Expand self-service only when the paved route proves safer and faster than bespoke delivery.
Engineering, architecture, security, risk, operations and domain product owners sign the release envelope.