101s.aiContact
Reference architectures

From guidance to wiring diagrams.

The joint guidance is principle-led and technology-agnostic, which is exactly why executives and architects need translation. The patterns below are concrete starting points that satisfy the guidance while staying recognisable to existing engineering teams.

Pattern 00

Agentic system anatomy

Before any pattern, agree on the components. An agentic system has, at minimum: an LLM reasoning core; a planning workflow that may spawn sub-agents; external tools the system can call; external data sources that flow into context; and short- and long-term memory. Every box is a potential attack surface.

AGENTIC AI SYSTEMUser / OperatorInput + oversightLLM Reasoning CorePlan · Decide · ActTriggers · Goals · PoliciesExternal ToolsAPIs · Code · Email · BrowseExternal DataRAG · Web · FilesMemoryShort / long-termPlanning WorkflowSub-agents · Delegation · TriggersSolid: outbound tool/action calls · Dashed: data flowing back into the prompt context
Components of an LLM-based agentic AI system. Each component is an attack surface.

Source: Careful adoption of agentic AI services, co-authored by ASD's ACSC, CISA, NSA, Canadian Cyber Centre, NCSC-NZ, NCSC-UK (Figure 1, p. 5).

Pattern 01

Zero-trust agent runtime

Three planes: a control plane that decides who/what/when; a data-execution plane where the agent actually runs in a sandboxed enclave; and a resource plane of approved tools and data. Every inter-plane call is authenticated, authorised, and logged.

CONTROL PLANETrusted Agent RegistryDIDs · mTLS certs · rolesPolicy Decision Pointper-request allow/denyJust-in-time Secretsephemeral · scoped · expiringAttestationcode/model hashDATA / EXECUTION PLANE — SANDBOXED ENCLAVEInput Managerprompt injectionfilter + sanitisetrust-tier on contextAgent (single principal)role-bound · least privilegenarrow tool allow-liststop-on-uncertainty defaultOutput Validatorschema · groundingcross-check redundant agentDLP on egressHITLapproveshigh-impactactionsRESOURCE PLANE — APPROVED TOOLS / DATAApproved APIRAG corpusRead-only opsWrite opsSub-agent
Zero Trust agent runtime: every action authenticated, authorised, logged.
Control plane

Identity & policy

  • Trusted agent registry — DIDs / mTLS certs, role bindings.
  • Centralised policy decision point — per-request allow/deny.
  • Just-in-time secrets — ephemeral, scoped, expiring.
  • Code attestation — agent must prove unmodified code.
Execution plane

Agent in a sandbox

  • Input manager: prompt-injection filtering, trust-tier on context.
  • Single-principal agent with narrow tool allow-list.
  • Output validator: schema, grounding, redundant-agent cross-check, DLP on egress.
  • HITL gate for high-impact actions.
Resource plane

Approved tools & data

  • Allow-list of pinned API versions.
  • RAG corpus with provenance metadata.
  • Read-only by default; writes via signed commands.
  • Sub-agent invocation gated by the policy decision point.
Implementation hints
  • Use OAuth2 token exchange + short-lived workload identity (e.g. SPIFFE/SPIRE) for the agent principal; never share long-lived credentials across agents.
  • Make every tool call go through an outbound proxy that enforces the allow-list and rate limits — the agent never reaches the public internet directly.
  • Run the agent process in a hardened sandbox (gVisor, Firecracker microVM, or equivalent) with seccomp + read-only root.
  • Log every prompt, tool call, model output to an append-only ledger separate from the agent's write scope.
Pattern 02

Multi-agent isolation with mediated handoffs

Implicit trust between agents is the most common cause of cascading agentic failures. Replace it with mediation: separation of duties (Orchestrator / Reader / Actuator), a policy validator, and a secondary reviewer agent that has to co-sign actions. High-risk agents live in their own enclave with a HITL gate.

ENCLAVE A · LOW RISKReader / RetrieverOrchestratorread-only · no write to logsMEDIATIONPolicy ValidatorSecondary Reviewer Agentconsensus · signed handoffsENCLAVE B · HIGH RISKActuatorHITL Gatecryptographically signed commandsRoles per CISA / OWASP: Orchestrator · Reader · Actuator with explicit boundaries
High-risk agents isolated; mediated handoffs replace implicit trust.

When to apply

  • You have more than one agent collaborating on a task.
  • Any agent can write to systems of record (CRM, ERP, email, code repos).
  • Compromise of one agent could pivot through trusted handoffs.

Anti-patterns to avoid

  • Agents sharing credentials or tokens.
  • Tool descriptions with persuasive language ranked by an agent during selection.
  • Sub-agent spawning without expiry timers or recorded grant chains.
  • Trusting another agent's output without schema validation or provenance.
Pattern 03

HITL workflow sized to reversibility

Not every action needs a human signature. Classify actions by impact, likelihood, and reversibility; set the gate accordingly. The agent does not get to decide which class an action falls into.

Planclassify impactLow risk · reversibleexecute · logMediummulti-agent consensusHigh · irreversiblehuman approval requiredQuarantine queuelog-deletes · resetsActioncryptographically signed
Human-in-the-loop checkpoints sized by reversibility and impact.

Low risk · reversible

Read-only retrieval, draft generation, internal summarisation. Execute and log.

Medium

Internal write actions, automated outreach, workflow updates. Require multi-agent consensus before execution; log the consensus chain.

High · irreversible

External payments, deletions, sensitive customer comms, network egress, system resets. Human approval — always — plus cryptographic signing on the resulting command.

Operational rule of thumb
Anything that cannot be reversed within a working day belongs in the high-risk tier and requires a named human signature. This includes any request to delete logs or audit records — quarantine those by default.
Pattern 04

Defence in depth, agent-shaped

Layer controls so that no single mechanism is load-bearing. Each layer below should fail before the next is breached. Treat this list as a checklist for any agent that touches production data.

AGENTSensitive data & critical systemsIdentity & least privilegeSandbox + isolation enclavesGuardrails & policy decision pointInput/output validationHuman-in-the-loop
Layered controls so no single mechanism is load-bearing.

Want a tailored architecture review?