[onexbot-occ-writer]

Summary

Adds a runnable successor for OMN-20332's generated behavior proof for omnidash#347. The generated item runs uv run pytest in omnidash. omnidash has no Python project, so pytest never starts, and the evidence sweep reported "Failed to spawn: pytest" (sweep comment aad76d02, 2026-10-02T04:30Z). Without this, the item stays red even after #347 merges.

What changed

• contracts/OMN-20332.yaml: one appended item, dod-occ-pr347-behavior-proof-runnable, with evidence_artifact: supersedes_dod_evidence:dod-occ-diff-derived-behavior-proof-pr-347. It uses the same shape as OMN-14756's successors. Its check runs the same test file the way omnidash's own CI does (.github/workflows/workflow-script-refs.yml): a bare environment with pytest and pyyaml, and the repo's pytest config detached.
  • uv run --no-project --with pytest --with pyyaml python -m pytest tests/ci/test_sibling_reads_pinned.py -q --noconftest -p no:cacheprovider -c /dev/null, cwd ${OMNI_HOME}/omnidash.
• drift/dod_receipts/OMN-20332/dod-occ-pr347-behavior-proof-runnable/test_passes.yaml: the executed receipt, bound by contract and entry hash.
• Nothing else is edited. The generated item stays, because merged entries are immutable.

How it was verified

• The check, run as written on omnidash#347's head 30541ab: 35 passed. Those are 31 per-workflow-file cases (counted with --collect-only), 2 extractor self-tests, the scanner's positive control and the required-checks.yaml check.
• The original command cannot start in omnidash. That is the sweep's own reported failure.
• The appended item parses as the 5th dod_evidence entry, and its folded check_value reads back as the exact command above.
• The receipt hashes were computed with omnibase_core's compute_contract_sha256 and compute_contract_entry_sha256 after yamlfmt re-wrapped the item. The entry hash was unchanged by the re-wrap, and the contract hash was recomputed.
• Pre-commit: every hook passes, including the receipt hardening gate.
• Not covered here: the item can only pass in the sweep once chore(OMN-7839): add dod_receipts for ModelConcurrencyContractSpec #347 is merged, because ${OMNI_HOME}/omnidash is the dev clone and the test file reaches dev with chore(OMN-7839): add dod_receipts for ModelConcurrencyContractSpec #347. The three ac-falsifier-acN failures come from the ticket's own falsifier text (uv run pytest …). They are not contract items and are not changed here.

Failure paths

n/a: this change writes no receipt, verdict or report beyond the declared evidence receipt above.

Open defects

none

Local gates

• No gate was bypassed — no --no-verify, no SKIP=, no --no-gpg-sign, no core.hooksPath override. Every pre-commit and pre-push hook ran.
• Any hook that failed was fixed at the input, not worked around.

Not in this change

• The ticket's AC falsifier lines. Making them runnable means editing the ticket text and re-accepting AC1–AC3, which is a separate decision.
• omnidash#347 itself, which is merged by Jonah.

[Patel230]

Evidence close-out for OMN-20332, read live at 2026-10-02T08:09:41Z.

onex_change_control#12281 merged as 23a7526dfab2308fede36792cffbd3cdd0b5e635, and current origin/dev contains it. It adds the runnable successor for omnidash#347's behavior proof (the bare-environment pytest form omnidash's own CI uses) and re-accepts AC1 to AC3 against it, with AC2 carrying the prefix control that must fail on the pre-change workflows.

The live CI rollup is 89 success, 2 skipped, 0 failed. The PR has zero comments and zero reviews.

What remains: the successor runs the test file from the omnidash dev clone, so it passes in the sweep only once omnidash#347 merges. #347 is still open.

Thank you!