Forge Intelligence

Forge Intelligence vocabulary

Canonical terms for orchestration, lmeta, LCDL overlays, and session dual-wiki.

Updated

Term ID pattern Meaning
Intelligence general.* / life.* Pack / cognitive composite
Capability capability.<intelligence>.<name> Atomic governed function
Intelligence node same as capability Runtime/DAG alias
Artifact artifact.<…> Typed payload
Artifact envelope pack artifact-envelope schema Cross-node handoff wrapper
lmeta unit unit.<intelligence>.<name> One node's declarative flow
lmeta library flow flow.<intelligence>.* Pack-level unit composition
LCDL contract intelligence_*_v1 FI-local executable I/O
Task-graph node T1 Scheduled capability in a case
Session sessions/<id>/ M/H dual-wiki run

Rule: 1 capability ↔ 1 lmeta unit ↔ 1 LCDL contract ↔ declared input/output artifacts.

Maturity axes (schema 1.1)

Every capability has two orthogonal axes in registries/capability_maturity.yaml. They are independent: high technical maturity does not imply high autonomy.

Technical axis (technical)

Implementation and runtime readiness of the intelligence node.

Level Meaning
I0_scaffold Unit + contract stub; envelope emit only
I1_contract Schema-valid contracts; not yet runtime-tested
I1_runtime Allowlisted; envelope/adapter runtime (contract evidence; tests green). Not domain assurance.
I2_integrated Wired into ≥1 multi-node case with fan-in; freeze gate green
I3_monitored Production attach + monitoring/reopening fields required
I4_assured Independent assurance / external audit evidence

Legacy field status is an alias for technical during migration; tools write technical only.

Evidence class (evidence_class)

What kind of evidence supports the current technical claim. Orthogonal to both axes above.

Class Meaning
fixture Scaffold / golden fixture only (default for I0)
contract Schema-valid FI contract + adapter runtime (default for I1_runtime)
semantic Domain validators + multi-node integration criteria met
expert Production monitoring / expert review evidence
assured Independent assurance / external audit

I1_runtime ≠ domain done: envelope/adapter tests prove contract compliance, not semantic correctness.

Lexicographic objective (C-11)

When policies conflict, FI applies safety → rights → correctness/evidence → traceability → resource cost. Token economy optimizes only after higher-priority gates pass.

Decision-autonomy axis (autonomy) — FI-local

How much the capability may run without a human gate. This is not the Forge L0–L8 execution autonomy ladder (blueprints-owned; agent delivery scope).

Level Meaning Human gate
A0_assisted Suggests / scaffolds; human approves emit or next step Continuous
A1_supervised Runs end-to-end; human reviews before downstream use Review-out
A2_bounded Auto within hard constraints; escalate on ambiguity/fail Exception-only
A3_monitored Bounded auto + monitoring / reopen triggers; audit trail Mission/policy only

Hierarchy rollups

registries/intelligence_maturity.yaml and MATURITY-HIERARCHY.md aggregate capability → intelligence → arena (general / life).

Disambiguation

Scheme Owner Meaning
FI technical I0–I4 forge-intelligence Node implementation maturity
FI autonomy A0–A3 forge-intelligence Decision-autonomy per capability
Forge L0–L8 blueprints Governed unattended agent execution scope
Pack maturity_axes I0–I4 Human-Life bundle Per-intelligence design implementation axis

See INTELLIGENCE-NODE-RUNTIME.md, NODE-CATALOG.md, and docs/domains/README.md.