The VCR roadmap's 39 artifacts are outside status_evidence_check's scope — two have no status key at all

[avrabe]

The roadmap's 39 artifacts are outside every status/evidence rule

scripts/status_evidence_check.py is the gate that enforces R0–R10 — a status
must be backed by evidence, evidence must belong to the release, a delivery
commit must be acknowledged, and so on. Its scope:

RELEASE_GLOB = "artifacts/release-v*.yaml,artifacts/release-v*/*.yaml,artifacts/release-v*/*.yml"

artifacts/verified-codegen-roadmap.yaml does not match. It holds 39
VCR-* artifacts — the entire North Star program, including every
proposed/approved/implemented status the roadmap advertises — and none
of them is subject to any of those rules.

Measured consequence

Two roadmap artifacts have no status key at all:

VCR-WCET-001   (no status)   Sound static per-function WCET bound (--emit-wcet)
VCR-WCET-002   (no status)   WCET bounded self-recursion via a verified depth-hint

Every other artifact in the repo has one. A missing status is not merely
untidy — it is unrepresentable in the lifecycle
(draft → proposed → approved → implemented → verified → accepted), and a
consumer filtering by status silently drops these two rows. rivet loads them
(they count toward ARTIFACT_FLOOR), so they are visible to the count and
invisible to the rules.

Why this is the same class as #1085 / #1091 / #1113 / #1119

Not "a missing gate" — the gate exists, is required, and is green. Its
scope just doesn't cover a file whose whole purpose is to record programme
status. That is the shape v0.61 was named for, one directory over:

• An artifact can be 'implemented' in vX.Y on evidence that shipped in vX.Y-1 — every gate passes and the release notes are wrong #1085 — evidence that could not fail on the failure it defines
• The traceability completeness gate has never fired: 60/60 release artifacts across v0.56-v0.60 have no linked sys-verification, and 'rivet coverage' prints instead of asserting #1091 — a gate that printed instead of asserting
• An oracle's ci-checks floor cannot see its decline half — 4 oracles could lose that half and stay green #1113 — a floor blind to half its own oracle
• status_evidence R4 is blind to conventional-commit delivery subjects — two v0.61 artifacts shipped and stayed 'proposed' with the gate green #1119 — R4 blind to a whole delivery-commit shape

The roadmap is arguably the worst place for it, because VCR-* statuses are
what the README's North Star table and the release planning both read from.

Not a simple glob widening

The release files and the roadmap have genuinely different obligations:

• Release artifacts carry release: and must satisfy R7/R8 (evidence belongs
  to the release, release: matches the filename) — the roadmap's artifacts
  mostly have release: None by design, since a programme item is not scoped
  to one release until it is.
• R4's delivery-commit rule assumes a release window; the roadmap has none.

So the fix is to decide which rules apply to programme artifacts — at
minimum: every artifact has a status, and the status is a legal lifecycle
value — and enforce those, rather than pointing the release-scoped ruleset at
a file it was not written for. Red-first is free: VCR-WCET-001 and
VCR-WCET-002 are an already-measured corpus of two, and a fix that does not
flag both has not fixed the reported thing.

Also worth deciding while here

39 roadmap artifacts carry release: None, including three the v0.62 planning
just had to scope by hand (VCR-TIER-001, VCR-REACH-002, VCR-RA-011).
Whether programme items should carry a target release once committed is a
separate question, but it is the reason this was noticed.