AGF

The 19 Primitives

Named patterns for governed agentic systems — runtime, lifecycle, and substrate primitives organized into the Rings Model.

The 19 primitives are the building blocks of AGF. They are not new inventions — they are named patterns drawn from established practice across distributed systems, security engineering, compliance, and control theory. The contribution is pulling them together into a composable architecture for the agentic context.

Runtime Primitives (#1–#17)

These operate within or across the ring pipeline during agent execution.

#PrimitiveRingOne-Line Description
1Separation of Producer & Verifier0+1The agent that creates output must not be the sole agent that validates it
2Validation Loops with Convergence Gates1Iterative verification until quality thresholds are met or budget is exhausted
3Self-Improving Cycles3The system gets better by learning from its own execution
4Adversarial Critique1Dedicated challenger that actively tries to find flaws
5Structured Output PersistenceFabricEvery ring produces and consumes structured artifacts — the data contract
6Provenance ChainsFabricEvery output carries its full decision history
7Bounded Agency2Agents operate within explicit, enforced boundaries
8Governance Gates2Explicit decision points where execution pauses for authorization
9Policy as Code2Governance rules as versioned, executable, testable objects
10Event-Driven ObservabilityFabricEvery ring emits structured events — the shared nervous system
11Trust Ladders2+3Trust is earned through demonstrated performance, not granted by default
12Memory-Augmented Reasoning0+3Agents reason with persistent memory, not just immediate context
13Error Handling & RecoveryFabricGraceful degradation, checkpointing, compensation
14Identity & AttributionFabricEvery agent has authenticated, inspectable identity
15Adversarial RobustnessSecurityAssume breach. Defense in depth. Verify explicitly.
16Transaction & Side-Effect Control0+2Pre-commit/commit/post-commit for irreversible actions
17Data Governance & Confidentiality2+FabricClassification, consent, PII, lineage, retention at every data flow

Lifecycle Primitive (#18)

#PrimitiveScopeOne-Line Description
18Evaluation & AssurancePre-deploymentThe gate before the gate — validates configurations before they reach production

Substrate Primitive (#19)

#PrimitiveScopeOne-Line Description
19Agent Environment GovernanceAll ringsGoverned composition of context, instructions, tools, workspace, and memory

How They Map to the Rings

The Rings Model showing primitive placement
  • Ring 0 (Execution): #1, #5, #12, #14, #16 — the agent works, produces structured output, with identity
  • Ring 1 (Verification): #1, #2, #4 — separate verifier, iterative loops, adversarial critique
  • Ring 2 (Governance): #7, #8, #9, #11, #16, #17 — boundaries, gates, policy, trust, transactions, data governance
  • Ring 3 (Learning): #3, #11, #12 — self-improvement, trust calibration, memory
  • Fabric: #5, #6, #10, #13, #14, #17 — structured output, provenance, events, errors, identity, data governance
  • Security: #15 — adversarial robustness as a pervasive concern

Primitive Interaction Tensions

The primitives are not always harmonious. AGF names seven tensions with architectural invariants that resolve them:

TensionConflictInvariant
Self-Improvement vs. Reproducibility#3 changes behavior; #6 requires traceabilityAlways reproducible at a specific version
Trust Ladders vs. Governance Gates#11 reduces oversight; #8 ensures accountabilityCan skip a spot-check, never regulatory approval
Bounded Agency vs. Self-Improvement#7 constrains; #3 improves — can it expand itself?The box can get smarter inside. It cannot grow itself.
Validation vs. Latency/Cost#2 improves quality; each pass costs time and moneyQuality bounded by economics, not just capability
Memory vs. Signal-to-Noise#12 accumulates knowledge; stale memories inject noiseGrowth without curation is hoarding, not learning
Policy as Code vs. Self-ImprovementCan Ring 3 change Ring 2's policy rules?Can suggest governance changes, cannot enact them
Environment vs. Governance Integrity#19 optimizes the environment — but it IS the control surfaceThe environment can get better. It cannot get less governed.

Gate Vocabulary & Governance Decision Records

AGF's gate-related concepts form a four-part vocabulary — not alternative spellings of one concept. See DECISIONS.md #8 for the full disambiguation.

  • Ring Control Signals (PASS / REVISE / HALT / GATE / DELEGATE / ERROR) — how primitives in adjacent rings signal each other via the Composability Interface. Emit observability events (Primitive #10), not GDRs.
  • Gate Resolutions (APPROVE / REJECT / MODIFY / DEFER / ESCALATE) — the Primitive #8 enum returned by an authorizer when a gate fires; also the return value of the GATE Ring Control Signal. Emit a Governance Decision Record (GDR).
  • Domain Outcomes — domain-specific gate decision enums (e.g., Tool Gate's Authorized / Conditionally Authorized / Denied) that map onto Gate Resolutions. Emit a GDR.

Every Gate Resolution and every Domain Outcome emits a Governance Decision Record — AGF's canonical audit artifact at gate boundaries. The GDR is the machine-readable companion to the human-readable rationale (dual-form principle, DECISIONS.md #5). The corresponding gate_resolved observability event references the GDR by decision_id.

Primitive Attribution

AGF's position is that the primitives are named patterns drawn from prior art, not novel inventions. The contribution is composition — pulling these patterns into a coherent architecture for the agentic context, naming their relationships, and specifying the invariants that let them work together. This table shows where each primitive's intellectual lineage sits.

#PrimitivePrimary lineageAGF contribution
1Separation of Producer & VerifierDistributed systems (code review, formal verification); CSA MAESTRO; NIST SP 800-53 SA-11Naming the producer/verifier split as a first-class agentic invariant
2Validation Loops with Convergence GatesControl theory; agentic research (reflection, ReAct); NIST AI RMF "Measure"Adding explicit budget/convergence gates to the verification loop
3Self-Improving CyclesReinforcement learning; MLOps continuous training; meta-learning researchRing 3 placement + tension invariants with reproducibility and bounded agency
4Adversarial CritiqueRed teaming; OWASP Top 10 for LLM Apps; adversarial robustness researchDedicated Ring 1 challenger role distinct from the verifier
5Structured Output PersistenceEvent sourcing; data contracts; OpenTelemetry semantic conventionsRing-to-ring data contract as Fabric primitive
6Provenance ChainsSLSA provenance; W3C PROV; SOC 2 / EU AI Act Art. 12 traceabilityExtending provenance to agent decisions, not just artifacts
7Bounded AgencyCapability-based security (POLA); sandboxing; NIST AC controlsApplying capability discipline to agent action surfaces
8Governance GatesApproval workflows; COBIT; ISO 9001; ITIL change managementUnified gate primitive + Gate Resolution enum + GDR artifact
9Policy as CodeOPA / Rego; GitOps; NIST CSF; CSA AICMSpecifying policy-as-code as a governance-ring primitive
10Event-Driven ObservabilityOpenTelemetry; SIEM practice; DORA researchMandating event emission at every ring boundary
11Trust LaddersCSA ATF (Intern → Principal tiers); Microsoft AGT maturityCross-ring (2+3) placement tying trust to observed performance
12Memory-Augmented ReasoningRAG; MemGPT; vector-store researchGovernance invariants over memory curation and revision
13Error Handling & RecoveryFault tolerance (Erlang/OTP); SAGA compensation; NIST CSF Respond/RecoverRing-aware error contracts (Fabric primitive)
14Identity & AttributionSPIFFE/SPIRE; NIST SP 800-63; Non-Human Identity practiceIdentity as a Fabric primitive spanning all rings
15Adversarial RobustnessOWASP Top 10 for LLM; MITRE ATLAS; NIST AI RMF; CSA MAESTROCross-cutting Security primitive with defense-in-depth invariant
16Transaction & Side-Effect ControlDatabase transactions (2PC); functional purity; tool-use safety researchPre-commit/commit/post-commit pattern for agent actions
17Data Governance & ConfidentialityISO 27001; NIST SP 800-53 SC/MP; GDPR; EU AI ActData-governance invariants at every agent data flow
18Evaluation & AssuranceMLOps eval; FDA software assurance; HELM / BIG-benchPre-deployment lifecycle placement ("gate before the gate")
19Agent Environment GovernanceContainer isolation; capability security; agent sandboxing researchComposed governance of context + tools + workspace + memory

How to read the table: "Primary lineage" names the source communities or specific works that AGF draws from. "AGF contribution" names what we think is genuinely additive — usually placement, naming, or invariant specification, not new theory. Where AGF's contribution is framing or synthesis, the underlying ideas remain the intellectual property of the cited communities.

Implementation Priority

For a practical implementation roadmap organized by phase, see the AI Engineering Profile.

For the complete primitive catalog with full patterns, deployment modes, security architecture, and existing work mapping, see the canonical source.

On this page