Handbook

L4.3 — Capability with human-signed ADR

Status: Defined (vision gates)

Updated

Sub-level at a glance
Parent level L4 — Feature/component
Autonomous unit Capability with agent-drafted, human-signed ADR
Distinct assay evidence L4.1/4.2 evidence + signed ADR reference
Grade today a

Canonical policy, grades a-d, and promotion criteria: Autonomy levels (Blueprints). Per-project scoring: Autonomy maturity framework (Blueprints).

Gates and evidence

All gates and core evidence of the parent level (L4 — Feature/component) apply unchanged. Declaring this sub-level adds the distinct check:

  • L4.1/L4.2 evidence, plus a human-signed ADR reference for the agent-drafted decision

A run that fails the sub-level check may still be reported at the plain parent-level grain if the parent's evidence holds — sub-level claims are strictly additive.

Prompt pack

Execution rubric (repo path, nav-hidden): prompts/autonomy/sublevel-l4-3.md — workcell-skeleton prompt declaring this sub-level, its evidence-to-return shape, and human gates.

Worked case

Hypothetical case — L4.x remains vision; requires human-signed ADR beyond agent draft.

Field Value
Repo forge-platform (hypothetical)
Goal Ship new capability with agent-drafted ADR signed by architecture owner
Evidence produced Pattern: runs/campaigns/<campaign>/<item>/run-*/machine/ with L4.1/4.2 evidence plus adr_signed_ref pointing to merged ADR
Assay pass rationale L4.3 _check_sublevel: non-empty adr_signed_ref after L4.1/4.2 evidence holds
Assay fail example Draft ADR only, unsigned → L4.3 evidence: signed ADR reference missing