Contracts
behavioral-contracts.md
Behavioural Contracts: plan guard/manifest vocabulary
No network/API interfaces; behavioural seams on existing in-process surfaces, each pinned by a test.
C-PLAN-1 — present spec.md reads present (FR-001)
Given a plan mission dir with committed spec.md, gather_artifact_presence(..., step_id="specify", mission_family="plan") yields present_artifacts ⊇ {spec.md} and _evaluate_plan_guards / evaluate_guards_strict return []. (Red pre-fix: ['Required artifact missing: spec.md'].)
C-PLAN-2 — manifest resolves the action vocabulary (FR-002)
_presence_filenames_for('plan') ⊇ {spec.md} and excludes the phantom goals.md; required_artifacts_for('specify','plan') == {spec.md}, ('plan','plan') == {plan.md}.
C-PLAN-3 — genuine block names its path (FR-003)
For a plan mission at specify with spec.md ABSENT, guard_failure_artifact_paths(...) returns a non-empty map keyed by spec.md.
C-PLAN-4 — cross-family parity (FR-004)
For each family in {software-dev, research, documentation, plan}, with the step's required artifact present on disk and presence gathered through the REAL gather_artifact_presence, the family's guard returns [] (no spurious block).
C-PLAN-5 — vocabulary-coherence guard (FR-005)
A corpus guard fails when a built-in mission type's manifest required_by_step keys diverge from its runtime step-contract action ids, or it is missing step contracts. Red against the pre-fix plan divergence, green after.
C-PLAN-6 — no non-plan regression (NFR-001)
software-dev/research/documentation presence filename sets and guard outcomes are byte-identical before/after (the per-(family,artifact_key) byte-compat characterization, minus the intentional plan rows).