Glossary
Canonical domain terms for Syndicate Code documentation.
Use this page when a term needs one stable meaning across public pages, proof records, and claim registry entries.
Direct answer
Runtime enforcement means controlling whether AI-proposed actions can execute, what evidence must be recorded, and what the product explicitly does not claim.
Claim classes
Verified claim
A public statement that is implemented and backed by a current claim, proof record, or validation artifact.
Defined in: canonical claim set
Do not use: implied, marketing-only
Qualified claim
A public statement that is true only within a bounded scope, partial implementation, or explicit exclusion set.
Defined in: canonical claim set
Do not use: absolute, universal
Product intent
A public statement about intended direction that is not a current guarantee.
Defined in: canonical claim set
Do not use: shipped, verified
Unsupported claim
A public statement that is not part of the canonical claim set and should not be presented as a product guarantee.
Defined in: canonical claim set
Do not use: claim, guarantee
Runtime enforcement
Policies, approval gates, and evidence rules that govern AI-proposed actions before side effects execute.
Defined in: runtime-enforcement, product, security
Do not use: monitoring-only, dashboard-only, safety blanket
Approval gating
A control-plane checkpoint that requires a review decision before a governed action can run.
Defined in: approvals, how-it-works
Do not use: notification, tooltip, advisory prompt
Audit chain
Append-only, hash-linked, HMAC-signed event sequence used as governance evidence.
Defined in: audit_and_evidence_model.md §2
Do not use: logs, history
Bounded autonomy
Workflow behavior constrained by policy, permits, checkpoints, and explicit boundary rules.
Defined in: product_system_definition.md §6 (G6)
Do not use: unrestricted autonomy, free-running execution
Canonical form
Deterministic normalized representation of a proposal used for policy evaluation and binding.
Defined in: governed_execution_spec.md §3
Do not use: canonical request text, normalized task
Checkpoint
Mandatory control-plane re-evaluation boundary requiring renewed approval, revalidation, or both.
Defined in: governed_execution_spec.md §3 and §7
Do not use: pause, confirmation prompt
Control plane
Authoritative enforcement component that evaluates policy, issues permits, and records evidence.
Defined in: product_system_definition.md §5
Do not use: server, backend, daemon
Governance-suspended mode
Runtime posture where interactive checkpoints are disabled, but policy evaluation, permit issuance, and audit writes remain active.
Defined in: Governance Modes
Do not use: policy bypass, audit-off mode, execution-unchecked mode
Delegation depth
Maximum allowed nesting level for specialist delegation chains within a governed workflow.
Defined in: autonomy_and_specialist_orchestration_spec.md
Do not use: unlimited delegation
Evidence
Durable records that support non-repudiation, completeness, immutability, and reconstruction.
Defined in: audit_and_evidence_model.md §2 and §5
Do not use: generic logs
Execution envelope
Bounded authorization scope that can allow multiple steps without per-step approval when policy allows.
Defined in: governed_execution_spec.md §3 and §6
Do not use: batch, task group
Fingerprint
Binding digest over permit-relevant context that prevents reuse outside approved scope.
Defined in: canonicalization_binding_and_approval_spec.md
Do not use: loose identifier
Governed execution
Lifecycle in which proposals pass canonicalization, policy evaluation, checkpointing, permit issuance, and evidence recording before side effects.
Defined in: governed_execution_spec.md §5
Do not use: autonomous execution, ungated automation
Operator
Human participant interacting through the TUI or CLI.
Defined in: syndicatecode-docs-site-spec.md §2.1
Do not use: user, developer, engineer
Permit
Control-plane-issued authorization object bound to approved canonical form or execution envelope.
Defined in: governed_execution_spec.md §3 and §8
Do not use: token, access token
Policy
Versioned authoritative rule bundle used by the control plane to allow, restrict, or deny behavior.
Defined in: policy_and_capability_model.md
Do not use: local preference config
Proposal
Structured request for work from operator, runtime, specialist, or workflow engine.
Defined in: governed_execution_spec.md §3
Do not use: task, request, command
Provider
Inference endpoint selected through policy-governed routing.
Defined in: provider_routing_and_model_abstraction_spec.md
Do not use: hardcoded model path
Session
Bounded interaction context for one operator and one workspace with explicit creation and terminal events.
Defined in: session_and_turn_model.md §3
Do not use: shell lifetime, process lifetime
Side-effect class
Risk classification for operations such as read, write, exec, network, secret, and delegate.
Defined in: governed_execution_spec.md §3
Do not use: permission bucket
Specialist
Bounded work unit with declared capabilities and constraints under governed execution.
Defined in: syndicatecode-docs-site-spec.md §2.1
Do not use: agent, subagent, assistant
Step
Single tool invocation within a turn; smallest auditable execution unit.
Defined in: session_and_turn_model.md ST-002a
Do not use: opaque action
Trust boundary
Independent dimension where trust is evaluated (tool, path scope, specialist, provider, workflow class).
Defined in: governed_execution_spec.md §3 and §7.2
Do not use: global trust bucket
Trust state
Per-boundary control state that determines whether execution can proceed with reduced friction.
Defined in: governed_execution_spec.md §3 and §7.2
Do not use: static trust score
Trust tier
Policy-configurable classification that adjusts approval friction without bypassing permits or evidence requirements.
Defined in: policy_and_capability_model.md
Do not use: trust level, permission level
Turn
Complete execution loop from one operator intent to terminal response or checkpoint.
Defined in: session_and_turn_model.md §5
Do not use: partial loop
Workflow
Named execution intent submitted through syndicate run and processed through governed execution stages.
Defined in: cli_command_reference.md §6
Do not use: ad hoc script run