plan(strategy): North Star + positioning as rivet artifacts (12 proposed, release-scoped)

[avrabe]

The plan made queryable rather than narrated. Twelve artifacts, all
proposed, nothing implemented and nothing claiming to be — a starting point
to fill in and recheck, not a finished roadmap.

Full audit behind it: https://claude.ai/code/artifact/a3753bd6-c1f3-415f-a1db-d35a0add16c9
Reproduced against f395518 (v0.40.0) on 2026-08-26; every requirement carries
the measurement that motivated it, with the command beside it.

Why this shape

release: is set on every requirement, so release readiness becomes a filter:

v0.41.0  REQ-CLAIM-LEDGER-001 · REQ-CLI-SURFACE-001 · REQ-SYSML2-CONFORMANCE-001
v0.42.0  REQ-HIR-FEATUREGROUP-PATH-001 · REQ-SYSML2-VISIBILITY-001   (+ existing WASM-GATES)
v0.43.0  REQ-FINDINGS-LAYER-001
v0.44.0  REQ-RENDER-COMPOUND-001 · REQ-RENDER-GOLDEN-001 · REQ-RENDER-METRICS-001
v0.45.0  REQ-SYSML2-FLOWS-MODES-001

That is the release-planning premise this repo already subscribes to and could
not previously exercise: 803 of 914 artifacts carry no release: at all, so
"is v0.X ready?" has not been answerable as a query.

The two decisions

DEC-NORTHSTAR-001 — the analysis engine for the SysML v2 + AADL world.

The ground moved in March 2026: Galois, with Ellidiss, CMU/SEI and Collins
Aerospace, released Systems-Modeling/SysML-v2-AADL-Release — a domain
extension library carrying most of AADLv2 core into SysML v2, inside the OMG
RTESC working group. That is the institutional core of AADL moving the language
into the SysML v2 ecosystem, on the back of OMG final adoption (July 2025) and
an official certification programme (June 2026).

Its stated exclusions are AADL flows and modes. Those are exactly what
latency, flow_rules, mode_reachability, modal and the whole
network-calculus arm consume. The standards track is carrying AADL structure
across and explicitly not the behaviour that makes it analysable — a narrow
opening nobody is contesting and one spar is already built for.

Reconciled explicitly with DEC-TSN-OSS-001, which names AADL as the
front-end/IR: this does not overturn it. That decision's own argument is that
the timing/NC math is a fabric-agnostic substrate paying for itself through
internal demand; this extends the same logic one layer up — the substrate is
formalism-agnostic too, so a SysML v2 front door adds an ingest path without
invalidating a single Tier-1 or Tier-3 requirement. Tier 2 keeps its kill-gates.

DEC-POSITIONING-001 — lead with the verdict, not the pipeline.

a compiler for system-architecture models → worst-case answers for real-time architectures. The compiler framing is kept, one level down: it
is an accurate answer to how this is built and it is why the codebase is as
good as it is. It fails as a headline for four measured reasons — a compiler is
judged on back-end fidelity and that is the weakest organ; it points at
translation when the value is judgement; its load-bearing word
"multi-formalism" is carried by the one front end with no outside oracle; and
it aims at a target that is receding now that formalisms are converging.

Also scoped there: drop the SysML v2 — KerML badge until the gate exists,
bound "replaces the Eclipse/Java OSATE2 toolchain" to what is measured (not
diagrams), and stop advertising SVG (Sugiyama layout) in the component
diagram — the README currently prints the renderer's root cause as a feature.

The evidence that drove the ordering

v0.41 buys nothing except the ability to know, which is why it is first:

$ cargo test -p spar-sysml2 --test conformance_tests
test parse_annex_a_simple_vehicle ... ok
test parse_package_example ... ok
… 8 passed; 0 failed

$ spar sysml2 parse test-data/sysml2/Package_Example.sysml
3 parse errors        # the same file the test above just passed

The assertion is assert!(!result.syntax_node().text().is_empty()) — it checks
the input source text is non-empty. Zero of the eight call errors().
The official corpus directories are empty, and grep sysml .github/workflows/*.yaml returns nothing.

That is the defect class REQ-GUARD-GATE-EVIDENCE-002 names, sitting in the
crate the strategy documents lean on hardest — four releases of that discipline
went to the CI guardrails and never to the tool's own interop claim.

Deliberately NOT in this PR

• No status demotions. REQ-SYSML2-PARSE is implemented and the audit
  contradicts it for the official corpus, but changing another artifact's status
  is a maintainer call, not a side effect of a planning PR. Flagged, not touched.
• No verification artifacts. Nothing here is built, so the right side of the
  V stays open. check_evidence_steps' stepless floor is unchanged at 78 —
  proposed requirements claim no evidence, which is the point.
• No README edit. DEC-POSITIONING-001 records the decision; executing it
  belongs with REQ-CLAIM-LEDGER-001 so the new wording lands already gated.
• The build-or-delegate call on the renderer is left open inside
  REQ-RENDER-COMPOUND-001 — it deserves its own decision once someone has
  costed extending etch against adopting an ELK-class engine.

Local verification

both files parse                                        OK
926 ids · 1255 links · 0 broken · 0 duplicates          OK
check_release_plane.py    exit 0   (status/release nested: 0)
check_human_scoped.py     exit 0
check_evidence_steps.py   exit 0   (floor 78, unchanged)

rivet validate is the CI gate; the local rivet is version-skewed and is not
trusted for a verdict here.

🤖 Generated with Claude Code

[github-actions]

Rivet verification gate

✅ 20/20 passed

	count
Passed	20
Failed	0
Skipped (no steps)	0

Filter: (and (= type "feature") (or (has-tag "v093") (has-tag "v0100")))

Failed artifacts

(none)

_{Updated automatically by tools/post_verification_comment.py. Source of truth: artifacts/verification.yaml.}