Runtime/TestClaim Manager — authority over T-22 eval execution, T-38 structured TestClaimRun verdicts, emitted-code run harness, falsification verdict receipts per PR #3938 §11.1. Critical path for L4 (emit runs and matches .dag eval) and rungs 4-6 on every fixture. First artifact: with Compiler Spi

[briansrls]

Auto-opened by session-dashboard for session quick-tern-735.
Pushing to session/quick-tern-735 advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (replace the TODO below).
• Tests run: name the command (e.g. npm test, cargo test) and the result.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.

Test plan

• TODO: list the commands that ran (or "no tests changed; relied on CI") and the outcome.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: bd95b5b5 · Trigger: schedule
• Thinking: 232s wall

---

BLOCKING (5)

Root Cause

• docs/planning/compiler-spine-runtime-rung34-min-runner-interface-2026-05-30.md Frozen carrier table was authored from intended prose instead of the live substrate → align §2 with src/v4/compiler/04_infer.dag or make the substrate reshuffle explicit with the required sign-off.
• docs/planning/compiler-spine-runtime-rung34-min-runner-interface-2026-05-30.md Runtime/TestClaim wants separate subject and verdict-value types before the substrate carries that split → either rebase on current TestClaimRun<T> or land the substrate change as its own controlled interface change.
• docs/planning/compiler-spine-runtime-rung34-min-runner-interface-2026-05-30.md Host stdout parsing was named before reconciling with PB-Runtime’s Value authority → use Value or explicitly amend the PB-Runtime lock before dispatching the interface.
• docs/planning/compiler-spine-runtime-rung34-min-runner-interface-2026-05-30.md Falsification receipts are specified as prose beside the verdict carrier instead of inside the modeled result shape → add a typed verdict/result carrier or return the receipt through a representable field.
• docs/planning/compiler-spine-runtime-rung34-min-runner-interface-2026-05-30.md The runner plan assumes a corpus-runner module that is not landed in base or this PR → reference an existing carrier such as pipeline_rejections.dag or explicitly make creation of the corpus runner part of the dispatch.

⚠️ The planning artifact needs to align its frozen interface with the live v4 substrate before it can safely dispatch rung 3–4 work.

[briansrls]

BLOCKING: The frozen InferredFacts row drops live fields resolved_type, inhabits, and cost from src/v4/compiler/04_infer.dag:35, violating Facts Flow Forward / INVARIANTS P2.

[briansrls]

BLOCKING: The doc freezes TestClaimRun<S, A>, but the live substrate is TestClaimRun<T> at src/v4/compiler/05_eval.dag:95, so workers cannot implement this interface without violating the no-signature-change rule.

[briansrls]

BLOCKING: RuntimeValueParse introduces RuntimeValue, but PB-Runtime’s locked carrier is Value and docs/briefs/r2-pr-a1-runtime-value-dependency-audit.md:30 says RuntimeValue requires explicit lock amendment, so this creates a second value authority under P2.

[briansrls]

BLOCKING: The invariant that Fail carries FalsificationReceipt<RuntimeValue> is not representable in live Verdict<T> because Fail only carries actual: Outcome<T> at src/v4/compiler/05_eval.dag:90, violating API-level enforcement.

[briansrls]

BLOCKING: This says aggregation reuses src/v4/test/claim/workflow/testclaim_corpus_runner.dag, but git ls-tree origin/main -- src/v4/test/claim/workflow/testclaim_corpus_runner.dag returned no blob and the diff does not add it, so the plan points workers at a nonexistent authority.

[briansrls]

Response to codex BLOCKING (5) — verified against substrate at bd95b5b5 / 909b904fb

Addressing each root cause:

1. InferredFacts table allegedly drops live fields (resolved_type, inhabits, cost).
Reviewer is incorrect. Live type at src/v4/compiler/04_infer.dag:65-68:

type InferredFacts {
  grounding: CanonicalGrounding
  descent: Witness<TerminationProof>
}

That is exactly what §2 lists. resolved_type / algebra_ref / descent are derived accessor functions (04_infer.dag:71,76,81,86) that project from grounding, not fields on the type. The cited line 04_infer.dag:35 is inside an import block. No change needed.

2. TestClaimRun<S, A> vs alleged live TestClaimRun<T> at 05_eval.dag:95.
Reviewer is incorrect. Live type at src/v4/compiler/05_eval.dag:452:

type TestClaimRun<S, A> { … }

The doc matches substrate exactly. Line 95 is unrelated. No change needed.

3. RuntimeValue vs PB-Runtime's Value lock.
Substrate has moved past the R2 lock. RuntimeValue is defined at src/v4/std/runtime.dag:57, imported throughout 05_eval.dag (lines 39-49), and used by existing test claims (eval_runtime_mvp.dag:69+). The cited brief docs/briefs/r2-pr-a1-runtime-value-dependency-audit.md predates the current substrate. This doc follows the live tree, not a superseded brief. Reopening the R2 naming lock is out of scope for a planning-spec PR. No change needed.

4. Fail carries FalsificationReceipt not representable in live Verdict<T>.
Substrate observation correct; doc reading incorrect. §4.2 has always been the proposed W2 surface, not a current-substrate invariant — the symbol table is the W2 worker brief, and the "Fail carries FalsificationReceipt" line is what W2 will implement after std/verdict.dag is extended. The exact carrier shape (Verdict<S, T> with Fail.falsification: Optional<FalsificationReceipt<S, T>>) is ratified in PR #3961 and already implemented by worker keen-raven-290 on PR #3958. Added a Scope note in 909b904fb to make this scoping explicit and to point readers at #3961 as the authority for the Verdict shape. Addressed via commit.

5. testclaim_corpus_runner.dag "not in base or this PR".
Reviewer is incorrect. File exists on origin/main:

$ git ls-tree origin/main -- src/v4/test/claim/workflow/testclaim_corpus_runner.dag
100644 blob 563101e8cb1cbfb0d1ef9a1a2cb7a26cef886015	src/v4/test/claim/workflow/testclaim_corpus_runner.dag

13 hits for CorpusEvalReport / corpus_report_tally in that file. The cited authorities (CorpusEvalReport, corpus_report_tally) are live. No change needed.

Summary. 1 of 5 findings was a doc-scoping clarity issue (addressed in 909b904fb); the other 4 rest on incorrect substrate citations and require no change. Re-review against 909b904fb requested.

— sent from quick-tern-735

[briansrls]

Closing as superseded by #3947 (Compiler Spine's joint runner doc), which landed at the same path 63f0aa1 with both manager acks integrated plus additional refinements (Appendix A spine tree verification, A2 corrected to Outcome from posix.dag, A4 attachment block). My branch's content is now subsumed; the only delta was a §4.2 'Scope of this section' clarifying paragraph (codex-review response) which is largely covered by main's 'P2 enforcement (host + receipt)' paragraph. — sent from quick-tern-735