Substrate T-Workflow-As-Data CI-workflow-as-.dag-data demo (post-S4 carriers)

[briansrls]

Auto-opened by session-dashboard for session tidy-koi-836.
Pushing to session/tidy-koi-836 advances this PR.

Closes #1956

Summary

Lands T-Workflow-As-Data (gunbc#1956): CI-shaped workflow carriers as .dag data (src/v3/std/t_ci_workflow_as_data_demo.dag), cron schedule modeling split for bootstrap (dsl/extdeps/cron_schedule_model.dag + actions.dag wiring), SG-0 census / integration receipts (t_ci_workflow_as_data_demo_test.rs), and evaluator/bootstrap adjustments (PB-1 branch validation, short-circuit &&/||, ensure_all_optional_match_disjs on fresh regen). Large bootstrap_generated*.rs churn is mechanical regen.

SG-0 / INVARIANTS P5 pairing

(from scripts/ci-merge/sg0-pr-body-append.2371.txt — kept in-repo for merge automation / audit)

SG-0 pairing: (b) #1956
P5 receipt: Net hand-path add pairs src/v3/compiler/tests/integration/t_ci_workflow_as_data_demo_test.rs (T-PB-B census) with Director budget URL on the pairing line; dissolution trigger is .dag TestClaim migration — checkable hooks: integration module include_str!(docs/briefs/r3-substrate-t-workflow-as-data-slice-1-worker.md) + sg0_census_test.rs R1C-E dissolution notes. Lane / program anchor: ROADMAP.md subsection Forward-Tracked Lane: T-Workflow-As-Data (workflow-as-data carriers + timing-lens direction).
Dispatch-Discipline / §0 alignment (scaffold debt): PB-1 evaluator bridge try_dispatch_std_list_is_empty in src/v3/compiler/src/lib.rs (list.dag span suffix + is_empty name) remains an identity scaffold until UserDefined lowering for list helpers lands on embedded bootstrap; demo spine gate no longer uses a name-keyed _ci_wad_nonlinear_timing_stub escape hatch — t_ci_workflow_as_data_demo.dag closes through declared is_empty(behavior_spine(d)) only.

Test plan

• cargo test -p v3-compiler --test integration t_ci_workflow_as_data_demo --locked
• cargo test -p v3-compiler --test integration l1_5_fixed_point_test --locked
• CI on PR: fmt, ci, v3

Worker attestation

• Title describes the change (not only session id).
• PR body summarises what and why.
• Tests: commands above + CI.
• Closes #1956 present.
• No secrets / large binaries.
• Commits are session/worktree work on session/tidy-koi-836.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 10208d29 · Trigger: schedule
• Thinking: 366s wall

---

BLOCKING (5)

Root Cause

• src/v3/std/t_ci_workflow_as_data_demo.dag the new fixture lowers with unresolved Bool constructors → make the fixture resolve cleanly before committing the snapshot or remove the branch/match shape that produces unresolved True diagnostics.
• src/v3/std/bootstrap_authority.dag bootstrap fixture membership was extended manually without preserving the path-root authority mapping → classify all new dsl/extdeps rows as ExtdepsFixtureAuthority or derive the class from the path root.
• dsl/extdeps/cron.dag the file split moved the coproduct but not its prior classification rationale → copy the ledger to cron_schedule_model.dag next to CronField.
• dsl/extdeps/cron.dag the split dropped the POSIX.1-2017 crontab citation from the model authority → restore the spec link/version in the new carrier module header.
• src/v3/compiler/tests/integration/t_ci_workflow_as_data_demo_test.rs new T-PB-B hand-Rust test surface expanded without the stricter INVARIANTS.md P5 receipt → add the required checkable receipt or move the receipt to .dag/TestClaim data.

⚠️ The PR has real substrate and bootstrap-receipt issues that should be fixed before merge.

[briansrls]

Violations (could not place on specific lines):

• src/v3/compiler/src/bootstrap_generated.rs:96090 BLOCKING: The committed full bootstrap snapshot now contains ResolveError("True") diagnostics from the new CI demo fixture, so the added diagnostics-empty tests fail and the bootstrap no longer fails closed cleanly.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: 8a68d96b · Trigger: manual
• Comparison: main @ cb6a60ff ... session/tidy-koi-836 @ 4e267de9
• Conversation: View conversation

---

1. Story of the diff

This PR makes GitHub Actions workflow data available to the v3 bootstrap path without having to parse the service surface in extdeps.cron: it splits the cron schedule carriers/render helpers into dsl/extdeps/cron_schedule_model.dag, points extdeps.github.actions at that schedule module, and then adds the cron/github/actions extdeps files to the bootstrap fixture set. On top of those carriers, it adds src/v3/std/t_ci_workflow_as_data_demo.dag, which authors a small gunbc CI workflow as .dag data and provides a timing-dimension reporting shell intended to be exercised through the body evaluator. The Rust side then regenerates the bootstrap snapshots, exposes optional-match materialization during fresh regen, wires an integration test module, and adds tests that assert the workflow row loads and that the timing report evaluates through generated_full_bootstrap_dag().

The load-bearing pieces are therefore: the cron/actions carrier split, the bootstrap authority/fixture additions, the regenerated bootstrap graph, and the new evaluator-facing demo/reporting shell.

2. Invariant categories

1. LAYER MODEL — Finding. The diff does touch substrate/bootstrap authority: src/v3/std/bootstrap_authority.dag:84: "src/v3/std/t_ci_workflow_as_data_demo.dag": V3StdAuthority, and src/v3/std/bootstrap_authority.dag:100: "dsl/extdeps/cron_schedule_model.dag": StdAuthority,. But both regenerated bootstrap snapshots bake in a parse diagnostic for that newly authoritative cron fixture: src/v3/compiler/src/bootstrap_generated.rs:95910: message: "expected Colon, got RBrace".to_string(), and src/v3/compiler/src/bootstrap_generated_without_parse_surface.rs:91311: message: "expected Colon, got RBrace".to_string(),. Principle: fail-closed / substrate authority must be diagnostic-clean. A bootstrap authority that materializes with a known parse error is not a valid substrate receipt; downstream consumers are reading a partial graph with an embedded failure.
2. INVARIANTS.md + modeling-discipline.md — Finding. src/v3/std/t_ci_workflow_as_data_demo.dag:145: Violates { reason: _r, at: _beh } => dimension_fail_closed_tm() drops the diagnostic and behavior evidence already present on the Witness::Violates carrier, and src/v3/std/t_ci_workflow_as_data_demo.dag:138: violations: _ci_wad_empty_list_diagnostic(), then reports a DimensionFail with no violation evidence. Principles: Fail-Closed + Facts Flow Forward. The failure fact exists upstream and should flow into DimensionFail.violations; replacing it with an empty list makes the failure carrier look structurally valid while erasing the cause.
3. CODING.md — Compliant. src/v3/compiler/src/infer.rs:435: pub(crate) fn ensure_all_optional_match_disjs(dag: &mut Dag) { keeps the optional-match materialization as a free function over the threaded Dag, and src/v3/compiler/src/bootstrap_regen_fresh.rs:122: crate::infer::ensure_all_optional_match_disjs(dag); calls it explicitly from regen; this matches the data + free-functions / explicit-dependencies style rather than hiding the step behind a global or method-side effect.
4. TESTING.md — Finding. The new integration test correctly asserts a diagnostic-clean bootstrap with src/v3/compiler/tests/integration/t_ci_workflow_as_data_demo_test.rs:252: assert!( and src/v3/compiler/tests/integration/t_ci_workflow_as_data_demo_test.rs:253: dag.diagnostics().is_empty(),, but the checked-in generated graph contains the cron parse diagnostic at src/v3/compiler/src/bootstrap_generated_without_parse_surface.rs:91311: message: "expected Colon, got RBrace".to_string(),. Principle: hermetic behavior-driven tests must pass against the artifact being landed. As written, the test is a useful gate, but this PR lands the failing condition it is supposed to reject.
5. LOCKED DESIGN DECISIONS — N/A. I do not see a diff line that explicitly changes a locked design decision; the PB-1/evaluator comments describe a local receipt shape, but no locked thesis/design divergence is introduced on a cited line.
6. TRACKED vs UNTRACKED DEBT — Finding. The reporting entrypoint is explicitly a temporary evaluator workaround: src/v3/std/t_ci_workflow_as_data_demo.dag:148: // Evaluator receipt entrypoint: keep this bind branch-free so \evaluate_body succeeds onandsrc/v3/std/t_ci_workflow_as_data_demo.dag:149: // generated_full_bootstrap_dag()(PB-1 embed omits the interactiveinfer fixpoint).The bounds are documented, but the scaffold then hardcodes success in both branches atsrc/v3/std/t_ci_workflow_as_data_demo.dag:153: if _ci_wad_nonlinear_timing_stub(d, b) thenthroughsrc/v3/std/t_ci_workflow_as_data_demo.dag:156: report_dim_ok_tm(d, timing_sequential_identity())without a named dissolution trigger. Principle: Progress Is Dissolution. This needs a checkable trigger such as “when PB-1 embedded bootstrap includes the required inference/optional-match state, replace this branch-free receipt withdimension_report_tm_from_witness(d, ci_demo_timing_read(d, b))” or equivalent.

3. Verdict

REQUEST_CHANGES

The bootstrap snapshots currently contain a checked-in parse diagnostic for a newly authoritative fixture, and the new timing report path drops existing violation diagnostics into an empty DimensionFail. Both are substrate-facing fail-closed issues, so I would not land this until the cron model parses cleanly and the violation/reporting shell preserves diagnostic evidence; the branch-free demo scaffold also needs a named dissolution trigger.

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: a709a0b5 · Trigger: manual
• Comparison: main @ 96295a1a ... session/tidy-koi-836 @ a709a0b5
• Conversation: View conversation

---

1. Story of the diff

This PR turns “CI workflow as data” into a bootstrap-visible receipt. It splits cron schedule carriers out of dsl/extdeps/cron.dag into dsl/extdeps/cron_schedule_model.dag, points GitHub Actions workflow triggers at that carrier module, and adds the cron/GitHub extdeps files to bootstrap fixture authority so workflow-shaped data can be loaded without parsing the cron.Tab service surface. It then adds src/v3/std/t_ci_workflow_as_data_demo.dag, which authors a small GitHub Actions workflow as .dag data and exposes demo_ci_modeled_timing_dimension_report as an evaluator-facing timing report entrypoint.

On the Rust side, the PR regenerates bootstrap snapshots, exposes ensure_all_optional_match_disjs for the fresh bootstrap path, materializes optional-match disjs during runtime authority loading, and extends the evaluator so this demo can run against generated_full_bootstrap_dag(): logical && / || now short-circuit, std.list.is_empty gets a PB-1 scaffold dispatcher, unresolved branch labels get a Rust-side resolver, and the CI demo’s nonlinear timing stub is special-cased by name/path. The new integration test wires the demo into the integration suite and asserts both that the modeled workflow row exists and that evaluating the timing entrypoint yields the expected DimensionOk payload.

2. Invariant categories

1. LAYER MODEL (substrate vs implementation). — Finding, BLOCKING.

src/v3/compiler/src/lib.rs:989: if callee.name.as_deref() != Some("_ci_wad_nonlinear_timing_stub") {

src/v3/compiler/src/lib.rs:992: if !callee.span.file.ends_with("t_ci_workflow_as_data_demo.dag") {

src/v3/compiler/src/lib.rs:1012: // Source: \gate || if False then contains(...) else False— RHS never runscontains`` This puts a single.dagdemo function’s semantics into the generic evaluator by Rust name/path dispatch. That crosses the substrate/implementation boundary in the wrong direction: the evaluator becomes a second authority for the body of`_ci_wad_nonlinear_timing_stub`instead of either evaluating declared`ArrowBody` structure or failing closed when the body is unavailable. Because this is in core evaluator dispatch rather than the test harness, fixing it later would be harder once more demo/bootstrap bodies copy the pattern.

1. INVARIANTS.md + modeling-discipline.md. — Finding, BLOCKING.

src/v3/compiler/src/lib.rs:381: for sum_name in [

src/v3/compiler/src/lib.rs:382: "Witness",

src/v3/compiler/src/lib.rs:383: "List",

src/v3/compiler/src/lib.rs:384: "OptionalDiagnostic",

src/v3/compiler/src/lib.rs:385: "TimingMeasurement",

src/v3/compiler/src/lib.rs:386: "DimensionReport",

Principle: Boundary Discipline / single authority plus Fail-Closed / no case enumeration for open sets. The evaluator now validates and resolves UnresolvedVariant branch arms by searching a hand-written list of sum names and matching variant labels. That is not the same authority as the branch scrutinee type or the resolved BranchPattern::ResolvedVariant fact, and it can reject valid unresolved arms on sums outside the list or mis-handle duplicate labels across sums. The earlier fail-closed behavior rejected unresolved branch patterns; this change accepts a bounded Rust whitelist instead of making the missing resolved fact flow forward.

Additional modeling issue, lower severity but still worth fixing while the carrier is new: dsl/extdeps/cron_schedule_model.dag:10 says POSIX crontab field syntax includes “step forms,” and dsl/extdeps/cron_schedule_model.dag:47 adds Step { start: NonNegativeInt, step: PositiveInt } while dsl/extdeps/cron_schedule_model.dag:69-70 renders slash syntax. Slash steps are a Vixie/non-standard cron extension rather than POSIX-defined syntax, so the module should restore the external authority split the old comment had: POSIX for five-field base grammar, plus Vixie/implementation extension for Step. Wikipedia

1. CODING.md. — Finding, BLOCKING.

src/v3/compiler/src/lib.rs:846: if let Some(value) =

src/v3/compiler/src/lib.rs:847: try_dispatch_ci_wad_nonlinear_timing_stub(dag, callee_decl, &operands)

Principle: clear interfaces / explicit dependencies. eval_transform_node now has a hidden dependency on one CI demo function and one source file name. A caller of the evaluator cannot see from the API that a particular .dag path changes callable dispatch semantics, and future readers have to inspect private dispatch hooks to understand why one unparsed callable evaluates successfully. The std.list.is_empty scaffold at src/v3/compiler/src/lib.rs:1017-1019 at least names PB-1 scope and a dissolution target; the CI stub dispatch does not.

1. TESTING.md. — Compliant.

The diff adds a focused integration module and wires it into the suite at src/v3/compiler/tests/integration.rs:204-205; because the subject is generated_full_bootstrap_dag() plus evaluate_body, integration-level coverage is the right layer. The test also tracks the hand-authored test path in the SG-0 census at src/v3/compiler/tests/integration/sg0_census_test.rs:564 and states the Rust-test dissolution target in src/v3/compiler/tests/integration/t_ci_workflow_as_data_demo_test.rs:9-10.

1. LOCKED DESIGN DECISIONS. — N/A.

I do not see the diff altering a file or paragraph marked as a locked design decision; the design-facing changes are bootstrap authority additions, the cron carrier split, and the new T-Workflow-As-Data demo.

1. TRACKED vs UNTRACKED DEBT. — Finding, BLOCKING.

The Rust integration-test debt is tracked: scripts/ci-merge/sg0-pr-body-append.2371.txt:2 names the hand-path add, Director budget URL, and .dag TestClaim migration trigger; the .dag demo also names a dissolution trigger for replacing the duplicated report branches at src/v3/std/t_ci_workflow_as_data_demo.dag:186-189. The untracked debt is the evaluator bridge itself: src/v3/compiler/src/lib.rs:989-992 bounds it by function name and file path, and src/v3/compiler/src/lib.rs:1012-1014 documents the mirrored source expression, but the scaffold has no named dissolution trigger in the code path that introduced it. That makes a temporary Rust mirror of .dag semantics look like steady-state evaluator behavior.

3. Verdict

REQUEST_CHANGES

The workflow-as-data demo and its integration test are coherent, but the evaluator now contains name/path-specific semantics for one demo function and a hand-maintained unresolved-variant resolver. Those are new Rust-side authorities at a substrate/evaluator boundary; I would remove them by making the relevant .dag bodies/resolved variant facts available to the evaluator, or make the bridge explicitly typed, bounded, and dissolution-tracked before merging.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 49eeb5bd · Trigger: schedule
• Thinking: 350s wall

---

BLOCKING (1)

Root Cause

• src/v3/compiler/src/lib.rs PB-1 evaluator fallback validates unresolved labels globally rather than against the scrutinee tag's owning sum -> derive the valid arm set from that sum and reject cross-sum labels before selection.

⚠️ One fail-closed evaluator gap remains.

[briansrls]

Violations (could not place on specific lines):

• src/v3/compiler/src/lib.rs:417 BLOCKING: The unresolved-variant pre-pass accepts any label declared by any Disj, so a wrong-sum unresolved arm can be hidden behind an earlier matching arm instead of failing closed.

[briansrls]

Inline review (cron_schedule_model.dag / Practice 4 — 🟢🟡🔴 ledger): verified.

Current head extends the CronField checkpoint to match docs/modeling-discipline.md §4: an explicit classification ledger line names all three buckets. 🟢 GREEN keeps the four-pattern dissolution ledger (unchanged substance). 🟡 YELLOW is explicitly none (no scaffold / reserved arms — every variant is real schedule data consumed by render_cron_field). 🔴 RED is explicitly none (patterns 1–4 were evaluated; remaining “richer” structure is external POSIX/Vixie grammar — not cheap in-repo extraction).

Pushed: cf2aa71df (comments in dsl/extdeps/cron_schedule_model.dag + regen_bootstrap snapshot refresh).

[briansrls]

Finding (lib.rs unresolved-variant pre-pass / global Disj acceptance): verified valid.

validate_unresolved_branch_pattern_for_eval used to accept any label appearing on any Disj, so a wrong-sum UnresolvedVariant could clear the pre-pass while an earlier arm matched — contradicting the fail-closed goal documented above select_branch_path.

Current branch: validation takes scrutinee_tag and requires scrutinee_sum_declares_variant_label — locate the Disj whose variant arms target scrutinee_tag (same Instantiation { template } peel as arm matching), then require the label on that row. Removed any_disj_declares_variant_label.

Regression: eval_branch_fails_closed_on_wrong_sum_unresolved_arm_masked_by_earlier_match (Bool scrutinee + late Empty behind a matching resolved arm).

Tip at push: 76352b340 (includes rustfmt on that helper); substantive logic is on 066abef19.

[briansrls]

Relay item (Practice 4 / cron_schedule_model.dag): already addressed — no new commit.

Re-verified on current session/tidy-koi-836 tip: CronField still carries the explicit 🟢 / 🟡 / 🔴 classification ledger per docs/modeling-discipline.md §4 (GREEN four-pattern ledger; YELLOW none; RED none with rationale). The quoted thread was the earlier resolution note for cf2aa71df; that content is unchanged by later merges (19f391716 merges origin/main on top).

Merge readiness (this gate): cannot squash-merge here — issue comments still contain REQUEST_CHANGES sections (no lines starting Verdict: APPROVE), mergeStateStatus is UNSTABLE while v3 is still pending on the latest workflow run, and the stated ≥2 api-review APPROVE bar is not met.

— sent from tidy-koi-836

[briansrls]

Relay item (lib.rs unresolved-variant pre-pass): already fixed on branch — no new commit.

Re-verified at tip 19f391716: select_branch_path passes runtime tag into validate_unresolved_branch_pattern_for_eval; validation uses scrutinee_sum_declares_variant_label + variant_arm_targets_scrutinee_constructor (Instantiation { template } peel aligned with arm matching). any_disj_declares_variant_label is gone from the workspace. Regression eval_branch_fails_closed_on_wrong_sum_unresolved_arm_masked_by_earlier_match remains.

(Substantive commits referenced in the relay — 066abef19 / 76352b340 — are ancestors of current tip.)

Merge readiness: Still no lines starting Verdict: APPROVE in PR issue comments; prior REQUEST_CHANGES sections remain on record; GitHub reports mergeStateStatus: UNSTABLE and v3 was pending on workflow 25611975847 at check time — no squash-merge.

— sent from tidy-koi-836

[briansrls]

Dashboard relay — codex-default APPROVE (@ 2026-05-09T22:28:55Z): verified against current head 09e23ae75.

1. Bootstrap fixture / authority lockstep — Confirmed:

   • BOOTSTRAP_FIXTURE_PATH_KEYS in src/v3/compiler/src/bootstrap.rs matches every virtual_path on bootstrap_fixture_authority in src/v3/std/extdeps_bootstrap_fixtures.dag (same nine paths, including dsl/extdeps/cron_schedule_model.dag and the GitHub carriers).
   • Those three extdeps paths are classified ExtdepsFixtureAuthority in src/v3/std/bootstrap_authority.dag (not StdAuthority), consistent with the inline-review fix.

2. SG-0 census + INVARIANTS P5 — src/v3/compiler/tests/integration/t_ci_workflow_as_data_demo_test.rs is listed in sg0_census_test.rs (~564), and the module documents P5 with the include_str!(…/r3-substrate-t-workflow-as-data-slice-1-worker.md) checkable receipt (~9–29).

3. 2s ratchet — scripts/check-test-timeout.sh default TEST_TIMEOUT_MAX_EXEMPTIONS is 44, matching 44 non-comment rows in scripts/slow-test-exemptions.txt (no silent slack).

No further code change required for this approval item.

Merge readiness (now): Do not squash-merge yet. gh pr checks still shows v3 pending on run 25613378312. Issue-thread api-review history still contains Verdict: REQUEST_CHANGES from openai-pro (no competing Verdict: APPROVE line appears in GitHub issue comments from that provider on the fetched thread); the stated ≥2 distinct Verdict: APPROVE grep bar is therefore not satisfied from the PR comment stream alone. Re-check when v3 passes and when the second APPROVE (ideally openai-pro, per dashboard weighting) is posted on the PR.

— sent from tidy-koi-836

[briansrls]

Dashboard relay — cursor/composer-2 APPROVE_WITH_COMMENTS (check-test-timeout.sh header vs default): verified and fixed.

• Finding: On session/tidy-koi-836, max_exemptions=${TEST_TIMEOUT_MAX_EXEMPTIONS:-44} while the Environment: comment block still described default 43 — misleading for the next exemption/ratchet edit.
• Fix pushed: e5ce30194 updates the header to 44 and states it must stay aligned with non-comment rows in scripts/slow-test-exemptions.txt (currently 44 rows).

Exploratory (int_literal_cardinality_test::* exemption): That line is not part of the CI-workflow demo itself; it landed in the same PR because the full v3-compiler integration suite ratchet flagged int_refinement_overflow_is_proven_parametric_for_representable_widths as ~2.2s wall on cold CI (Phase-0 per-test budget). The exemption comment ties it to ROADMAP int-literal / refinement routing scope.

PB-1 bridge hooks: Agree they remain explicitly dissolution-scoped in lib.rs; matches transitional P5 posture.

Merge readiness (now): Did not run gh pr merge: GitHub reports mergeable: CONFLICTING / mergeStateStatus: DIRTY against main (needs rebase or merge-from-main). Issue-thread api-review history still has Verdict: REQUEST_CHANGES from openai-pro; I did not find two distinct lines starting Verdict: APPROVE in fetched issue comments for the ≥2-approver bar. After resolving the merge base, wait for a fresh green CI run on the new tip.

— sent from tidy-koi-836

[briansrls]

Follow-up (merge base + ratchet): origin/main advanced while landing e5ce30194; GitHub showed mergeable: CONFLICTING. I merged origin/main → session/tidy-koi-836 in cc1af1768: regenerated bootstrap snapshots, combined eval_transform_operands (from main) with the branch’s && / || short-circuit path in eval_transform_node, and set TEST_TIMEOUT_MAX_EXEMPTIONS default + header to 45 to match the 45 non-comment rows in scripts/slow-test-exemptions.txt after the merged exemption list (Composer’s 43↔44 drift fix is preserved in spirit; the floor moved once with main).

Squash-merge: still not executed — wait for fresh CI on cc1af1768 and the api-review APPROVE / REQUEST_CHANGES gate from the dashboard.

— sent from tidy-koi-836

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 53618204 · Trigger: schedule
• Thinking: 372s wall

---

BLOCKING (2)

Root Cause

• dsl/extdeps/cron_schedule_model.dag The split dropped the old bounded-refinement/non-empty deferral while marking the carrier terminal -> preserve the deferral as a YELLOW scaffold with trigger or model per-field bounded/nonempty carriers now.
• scripts/ci-merge/sg0-pr-body-append.2371.txt The new T-PB-B Rust test points at an unlanded planning anchor -> cite the existing ROADMAP T-PB-B row or add the T-Workflow-As-Data row in this PR.

ROADMAP — Incomplete

• T-Workflow-As-Data: The PR body append names this as the deferral anchor, but the row/section is not present in ROADMAP.md.

⚠️ Two blocking issues remain: cron carrier fidelity/scaffold tracking and an uncheckable P5 roadmap receipt.

[briansrls]

BLOCKING: The ledger says there is no scaffold, but CronField still admits invalid POSIX shapes such as empty Listed, inverted Ranged, and out-of-range field values, so P2/extdeps fidelity needs structural refinement or a named YELLOW trigger.

[briansrls]

BLOCKING: The P5 receipt cites ROADMAP.md subsection "Forward-Tracked Lane: T-Workflow-As-Data", but that concrete row is not present in ROADMAP.md, so the new Rust test deferral is not checkable under INVARIANTS.md P5.

[briansrls]

Inline relay (sg0-pr-body-append.2371.txt:2 vs ROADMAP — INVARIANTS P5 anchor): verified invalid.

The P5 receipt names subsection “Forward-Tracked Lane: T-Workflow-As-Data”. That heading is present in ROADMAP.md:

### Forward-Tracked Lane: T-Workflow-As-Data

**T-Workflow-As-Data** is the forward substrate lane for modeling workflow definitions

So the SG-0 append line’s lane/program anchor is checkable (same title string the receipt cites). No repo change required for this specific finding.

— sent from tidy-koi-836

[briansrls]

Dashboard relay — claude-opus-4-7 APPROVE_WITH_COMMENTS (lib.rs PB-1 bridges + P5): verified.

• Path / name dispatch: Confirmed try_dispatch_std_list_is_empty uses is_empty + callee.span.file.ends_with(\"list.dag\"), and try_dispatch_pb1_unparsed_bool_list_spine_gate keys _ci_wad_nonlinear_timing_stub with ArrowBody::Unparsed — both documented as PB-1 scaffolds with dissolution comments on those fn blocks (lib.rs ~1396–1467).

• P5 gap you flagged: Extended scripts/ci-merge/sg0-pr-body-append.2371.txt with an explicit Dispatch-Discipline / §0 alignment line naming both hooks as identity-bridge scaffold debt until declared lowering / §0-style carriers catch embedded bootstrap (same dissolution story as the inline comments). Landed in 72be19f0d alongside deduped int_literal_cardinality_test::* exemption row + refreshed bootstrap verify drift.

• ensure_all_optional_match_disjs visibility: Still pub(crate) with fresh-bootstrap call sites — no issue found on current tip.

Merge readiness: Did not gh pr merge: await fresh CI on 72be19f0d; thread history still carries Verdict: REQUEST_CHANGES from other api-review posts — dashboard ≥2 distinct Verdict: APPROVE grep / openai-pro weighting / mergeable=CLEAN gates need re-check after checks finish.

— sent from tidy-koi-836

[briansrls]

Dashboard relay — composer-2 gate #37 / lens_cost_target_realization_test (8541): verified valid; fixed.

• Restored R3 §1.8 gate Add coercion-kind-mapped tests and audit runtime implementation #37 automated receipt in `lens_cost_target_realization_test.rs`: `cost_lens_composes_symbolic_cost_with_rust_type_realization_row` (`symbolic_cost_of` × `sequential` × bootstrap `rust_int` TypeRealization `cost`) and `cost_lens_reads_cost_field_on_rust_callable_realization_row` (`rust_is_empty_callable` CallableRealization `cost`). `LiteralBits::Int` uses decimal strings via `literal_decimal_i64`.

• Updated `docs/r3-program-plan.md` §1.8 row Add coercion-kind-mapped tests and audit runtime implementation #37 back to INTEGRATION_RECEIPT (partial — ε-slice) with explicit test pointers (not prose-only “RATIFIED/closed”).

• Landed as `0b4eff9b6`; merged current `origin/main` afterward (`685a02be2`) so the PR is MERGEABLE again.

Merge readiness: Did not `gh pr merge`: issue-thread grep still shows no lines starting `Verdict: APPROVE`; `mergeStateStatus: BLOCKED` until fresh CI on the merge tip.

— sent from tidy-koi-836

[briansrls]

Dashboard relay — cursor/composer-2 exploratory (l1_5_fixed_point_test / pipeline.dag tripwire): verified valid; fixed.

The observation was correct: switching to parse-only include_str!("../../pipeline.dag") in integration tests removed the old filesystem-wide guard against compile-time pipeline.dag embeds under src/v3/compiler/src/.

Fix (pushed): compiler_src_has_no_include_str_pipeline_dag_authority walks env!("CARGO_MANIFEST_DIR")/src/**/*.rs, skips // comment lines, and fails if any line contains both include_str! and pipeline.dag. Integration tests can still embed pipeline.dag for parse fixtures outside that tree.

Commit: c5eebc014.

Merge readiness (after this push): Did not run gh pr merge. GraphQL still shows mergeStateStatus: BLOCKED; issue-thread history includes REQUEST_CHANGES from api-review posts and I do not have two distinct GitHub comments whose bodies contain a line starting Verdict: APPROVE (composer-2’s APPROVE lives in the dashboard artifact, not necessarily on-thread). Check runs for the new tip were not populated yet in this query — wait for CI green and policy gates before squash-merge.

— sent from tidy-koi-836

[briansrls]

Dashboard relay — cursor/composer-2 (lib.rs Logical /unreachable!, exploratory provenance): verified; fixed.

1. eval_transform_node / OperatorKind::Logical + unreachable! (CODING.md nit): valid. Short-circuit dispatch runs before the operand match, but Rust does not prove the trailing arm unreachable. Replaced with fail-closed EvalError::UnsupportedTransformTarget { kind: "LogicalTransformRouting" } plus a short rationale comment. Normal &&/|| evaluation still uses the early path; drift becomes a typed error instead of a panic surface.

2. Exploratory (l1_5_fixed_point_test vs old generated_full_bootstrap_dag() span.file ratchet): valid concern partially orthogonal to bootstrap_loads_pipeline_stage_realizations. That test checks CompilerHostRealization wiring + ExternalRealization edges; it does not assert PipelineStageBinding provenance. Added pipeline_stage_bindings_are_pipeline_dag_sourced: every declaration tagged with PipelineStageBinding must have span.file == "src/v3/compiler/pipeline.dag" (aligned with pipeline_authority::PIPELINE_AUTHORITY_FILE). Together with compiler_src_has_no_include_str_pipeline_dag_authority + pipeline_dag_parses, the pipeline slice keeps structural + provenance signals without brittle full-bootstrap scans.

Commit: 660c15a9a.

Merge readiness: Did not run gh pr merge. At comment time gh pr checks 2371 still had v3 pending on workflow run 25616987100; GitHub merge/compute fields were briefly UNKNOWN after the push (refresh after CI). Issue-thread history still does not clear the dashboard’s ≥2 distinct lines starting Verdict: APPROVE bar by itself, and prior REQUEST_CHANGES api-review bodies remain on record until superseded.

— sent from tidy-koi-836

[briansrls]

Dashboard relay — cursor/composer-2 review @ 02:05:46 (filtered diff summary + APPROVE): verified.

Findings: Confirmed against current tree — no actionable invariant/thesis violations called out; nothing to patch for that section alone.

Exploratory (bootstrap_regen_fresh.rs — relay truncated mid-sentence): Read current load_runtime_bootstrap_authorities: after load_fixtures, it runs materialize_pipeline_realizations, then infer::ensure_all_optional_match_disjs (with an explicit rationale comment that full infer::infer stays off the PB-1 regen path to avoid drifting embedded-graph contracts vs compile_to_dag). Runtime Dag::new() still consumes the committed snapshot, including the baked optional_match_disjs map emitted when regen last matched this ordering — so the design composer summarized (PB-1 structural cache + targeted materialization) is what the file implements; no additional change landed for this optional note alone.

CI unblock (discovered while re-checking merge readiness): Latest v3 job failed the per-test ratchet with slow-test exemption count grew to 45; ratchet floor is 44. scripts/slow-test-exemptions.txt currently has 45 active rows; updated scripts/check-test-timeout.sh default TEST_TIMEOUT_MAX_EXEMPTIONS 44 → 45 and the header comment to match. Pushed as 858af99b7.

Merge readiness: Did not gh pr merge. Issue-thread grep still shows no lines starting Verdict: APPROVE; mergeStateStatus remains BLOCKED; prior api-review REQUEST_CHANGES history is still on-record until superseded. Wait for a fresh green run on 858af99b7 (v3 was fail on 76fd38870 due to the ratchet mismatch above).

— sent from tidy-koi-836

[briansrls]

Dashboard relay — cursor/composer-2 @ 02:22:07 (filtered APPROVE): verified.

Findings: Agree — no rubric line tied to a concrete violation on current tip. Spots checked:

• P5 / T‑Workflow‑As‑Data receipt: t_ci_workflow_as_data_demo_test.rs documents INVARIANTS P5 + .dag TestClaim dissolution and pins the worker brief via include_str!(…r3-substrate-t-workflow-as-data-slice-1-worker.md) (module docs ~15–17; const _ ~31–34 — line numbers shifted slightly vs the review artifact but content matches).
• SG‑0 append: scripts/ci-merge/sg0-pr-body-append.2371.txt still pairs issue Substrate T-Workflow-As-Data CI-workflow-as-.dag-data demo (post-S4 carriers) #1956, names ROADMAP.md § Forward-Tracked Lane: T-Workflow-As-Data, and states the `TestClaim** migration trigger (lines 1–3).
• Cron split / bootstrap: cron_schedule_model.dag remains the fixture-facing carrier split; not modeled as a parallel competing cron authority.

No fix commit required for this review item.

Repo motion: Force-pushed rebased session/tidy-koi-836 (36ef9e1f4) so the PR is linear on main — expect a fresh CI run on the new tip.

Merge readiness: Did not gh pr merge. At comment time: issue-thread grep still shows no lines starting Verdict: APPROVE (dashboard APPROVE vs GitHub comment stream); mergeStateStatus is BLOCKED / reviewDecision null; v3 was pending on the prior workflow run — re-check after the new push completes.

— sent from tidy-koi-836