Docs menu

Semantic domains

A semantic domain is a named meaning space — staging, pointer, sentinel, versioning, and so on. The Semantic Engine maintains a constraint graph over concepts in those domains. When an agent asserts something that conflicts with an existing concept (for example: “commits may use raw git while CAP is on”), semantic_enforce surfaces the contradiction before the commit.

Why it exists

  • Agents forget prior constraints within a long session.
  • Docs, rules, and code often encode the same idea three incompatible ways.
  • Governance needs a machine-checkable place to say “this concept belongs to domain X and implies Y.”

Core domains

Domain ids agents reason about text
# Core semantic domains agents reason about
staging | semantic | pointer | sentinel
versioning | deployment | canonical | data-array
efficiency | impact | policy | health
DomainOwnsDocs
stagingAtomic Commit-ScopesStaging scopes
pointerPath → home mapsPointer domains
canonicalSingular SoT per conceptCanonicalization
data-arrayOrdered / unique listsData arrays
versioningSemver bump policyVersioning
deploymentEnvironment promotionDeployment
sentinelHalt / remediate reflexesShowcase
efficiency · impactToken cost and outcome maturityEngine registry

Agent usage

semantic_enforce text
// MCP
semantic_enforce {}
// Failures name the conflicting concepts and suggested next tools.

Each engine in the registry declares its semanticDomain so hosts can filter tools and docs by meaning space.

AGS documentation