Repository navigation
evidence(OMN-15488): OCC DoD contracts + PASS receipts for OMN-15488 and OMN-15239 - #5759
Merged
Merged
Conversation
…l tickets still missing one (OMN-15488, OMN-15239) Authors contracts/OMN-15488.yaml and contracts/OMN-15239.yaml plus 17 PASS receipts under drift/dod_receipts/, closing the CONTRACT_MISSING gap that blocks the Steel dod_verify closeout for these two tickets. Scope correction vs the dispatch: the lane was scoped to FIVE missing contracts (OMN-15488, 15239, 15535, 15490, 15493). Live readback of onex_change_control origin/dev found three of them ALREADY landed in #5756 (9089ffa, contracts/OMN-{15535,15490,15493}.yaml). Only 15488 and 15239 were genuinely absent, so this PR authors two, not five. Re-verified per ticket with `git cat-file -e origin/dev:contracts/OMN-<n>.yaml`. Per-ticket binding (each contract pins its OWN merged PRs + merge SHAs, all on jonahgabriel/steel_onslaught main): OMN-15488 -> steel#235 @ff5df34fdb374ddcffb9d14576a16b49ebf1c87e steel#240 @a3b0d8aa7f7392942f6d78c47bf7f98247c63d09 (executing pin) steel#241 @fa2bd9d4d3d94c8bab3d3696e274ca002c284eac (evidence docs) OMN-15239 -> steel#225 @23a88d09051565270f76cade8bc7b685cd790f38 Binding trap handled explicitly: node_dod_verify's evidence_collector autobinds PR references against _DEFAULT_GITHUB_ORG="OmniNode-ai", which resolves nothing for a personal repo. Every check therefore carries the PR number and `--repo jonahgabriel/steel_onslaught` in the SAME shell clause, which is what the collector's `_hardcoded_pr_bindings_in_value` same-clause extractor requires. 17 checks, all hosted-runnable and pinned to immutable merge SHAs (never a feature-branch ref, never a workflow-run conclusion). Executed GREEN as written on the gate host both before and AFTER the yamlfmt reflow, then RED-proven: 14 flip RED at their pinned commit's own first parent. The other 3 (the OMN-15488 deliverable-1 source items) pin a3b0d8aa, whose first parent e3dbd017 ALREADY contains deliverable 1 -- it landed earlier in #235 -- so the naive first-parent oracle is vacuous for them; they are RED-proven against be4dbe49, the first parent of #235, and all three flip RED there. This is recorded in the contract summary, not just here. Honest scope notes carried in the contracts themselves: - OMN-15488 AC1/AC2/AC3/AC6 attest the independent gates agent's RECORDED results at an immutable SHA; they do not re-execute the gates, because the artifacts live in a local hermetic snapshot no hosted runner can reach. - AC2 was discharged by per-block invocations of check_contamination_gate.py, not one whole-lane run, because that script's interface cannot express a three-phase lane. Weaker than a single whole-lane exit-0; stated as such. - AC5's scoring document is titled DRAFT, and the battery's pre-registered primary landed NOT-SUPPORTED (a reported negative, per the ticket's own pre-declared bands) -- the checks pin the computed statistics, not a verdict word, so a favorable substitute statistic could not pass. Author is not the implementer: none of the underlying steel work was built by this lane (fable-steel-0731 wave-2). No-self-authored-evidence holds. No Linear status is changed here. Per doctrine a Done flip routes only through dod_verify with durable evidence. Cites OMN-15488, OMN-15239.
|
Warning Review limit reachedYou’ve reached a temporary PR review limit under our Fair Usage Limits Policy. Next review available in: 46 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (4)
📝 WalkthroughWalkthroughAdded enriched contracts and pinned DoD command receipts for OMN-15239 semantic retry behavior and OMN-15488 receipt evidence. ChangesOMN-15239 semantic retry evidence
OMN-15488 receipt evidence
Estimated code review effort: 3 (Moderate) | ~25 minutes Possibly related PRs
Suggested labels: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
…ts for both cited tickets occ-preflight/eligibility resolves the tickets a PR owns from the title tokens when the body carries no closing keyword, so this PR owns OMN-15488 and OMN-15239 and each needs a PASS receipt bound to PR #5759. Appends the standard occ-self-bind item + receipt to each contract, mirroring occ-self-bind-pr-5756 and occ-self-bind-pr-5544. Append-only: every pre-existing dod_evidence entry is byte-intact and all 17 original receipts still hash-match their contract entries (per-entry hashing, OMN-13888, which folds in only the receipt own entry plus the immutable header). Binding stamp only; flips no Linear status. Cites OMN-15488, OMN-15239.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@contracts/OMN-15488.yaml`:
- Around line 215-227: Strengthen the AC5 NOT-SUPPORTED verification by
replacing the bare token grep in contracts/OMN-15488.yaml lines 215-227 with an
exact check for the pre-registered band clause “NOT-SUPPORTED — D_ws <= 0.”;
apply the identical check_value change in
drift/dod_receipts/OMN-15488/dod-omn15488-ac5-primary-scored-as-prereg/command.yaml
line 7 and re-run the probe.
- Around line 192-206: Update the duplicated AC4 verification command so it
asserts the seed-4101 first promote-phase match and retains the budget-window
check, rather than checking the NO-PROMOTION phrase twice. Apply the identical
check_value change in contracts/OMN-15488.yaml lines 192-206 and
drift/dod_receipts/OMN-15488/dod-omn15488-ac4-promotion-within-budget/command.yaml
line 7, then re-run the probe to refresh the receipt.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 976c9872-c035-4617-94ab-32cc9d27b810
📒 Files selected for processing (19)
contracts/OMN-15239.yamlcontracts/OMN-15488.yamldrift/dod_receipts/OMN-15239/dod-omn15239-bare-semantic-error-no-longer-escapes/command.yamldrift/dod_receipts/OMN-15239/dod-omn15239-budget-shared-with-card-mode/command.yamldrift/dod_receipts/OMN-15239/dod-omn15239-dated-prereg-amendment/command.yamldrift/dod_receipts/OMN-15239/dod-omn15239-exhaustion-raises-catchable/command.yamldrift/dod_receipts/OMN-15239/dod-omn15239-pr-225-merged-at-23a88d09/command.yamldrift/dod_receipts/OMN-15239/dod-omn15239-repair-preserves-boundary/command.yamldrift/dod_receipts/OMN-15239/dod-omn15239-tests-exhaustion-and-recovery/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-ac1-prereg-timing-gate-exit-zero/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-ac2-ac3-ac6-verdict-table-all-pass/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-ac4-promotion-within-budget/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-ac5-primary-scored-as-prereg/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-d1-contract-tests/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-d1-genesis-knob-phase-caps/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-d1-overlay-mirror-stanza/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-pr-235-merged-at-ff5df34f/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-pr-240-merged-at-a3b0d8aa/command.yamldrift/dod_receipts/OMN-15488/dod-omn15488-pr-241-merged-at-fa2bd9d4/command.yaml
Comment on lines
+192
to
+206
| - id: "dod-omn15488-ac4-promotion-within-budget" | ||
| description: >- | ||
| AC4 — promotion fired inside the pre-registered promote budget of 15, so the alternative NO-PROMOTION | ||
| escape branch was not taken and, critically, no post-hoc budget extension occurred. At fa2bd9d4 | ||
| the scoring document records that `policy_promoted` fires on the FIRST promote-phase match, seed | ||
| 4101, against the declared `budget was 15 attempts, seeds 4101-4115`, and states in terms that `The | ||
| **NO-PROMOTION` escape does not trigger. Both legs are asserted: firing on the first attempt is | ||
| the fact, and the budget window being the pre-registered 4101-4115 (not a widened range) is what | ||
| forecloses the post-hoc-extension failure mode the ticket explicitly pre-declared against. | ||
| source: "manual" | ||
| status: "verified" | ||
| checks: | ||
| - check_type: "command" | ||
| check_value: |- | ||
| s="$(mktemp)" && gh api 'repos/jonahgabriel/steel_onslaught/contents/docs/evidence/2026-07-31-lgate2-decisive-battery-scoring.md?ref=fa2bd9d4d3d94c8bab3d3696e274ca002c284eac' --jq .content | base64 -d > "$s" && grep -qF 'NO-PROMOTION escape did not trigger' "$s" && grep -qF 'The **NO-PROMOTION' "$s" && grep -qF 'budget was 15 attempts, seeds 4101-4115' "$s" && rm -f "$s" |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick win
AC4 asserts a promotion leg that its command does not check. One command string is duplicated in the contract and in its receipt. Both copies assert the NO-PROMOTION negative twice and the budget window once, and neither asserts that policy_promoted fires on the first promote-phase match at seed 4101, which the contract description claims.
contracts/OMN-15488.yaml#L192-L206: replace the duplicategrep -qF 'The **NO-PROMOTION'leg with a leg that pins the seed-4101 firing sentence, or narrow the description to the legs actually asserted.drift/dod_receipts/OMN-15488/dod-omn15488-ac4-promotion-within-budget/command.yaml#L7-L7: apply the identicalcheck_valueedit and re-run the probe so the receipt records the corrected command.
📍 Affects 2 files
contracts/OMN-15488.yaml#L192-L206(this comment)drift/dod_receipts/OMN-15488/dod-omn15488-ac4-promotion-within-budget/command.yaml#L7-L7
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@contracts/OMN-15488.yaml` around lines 192 - 206, Update the duplicated AC4
verification command so it asserts the seed-4101 first promote-phase match and
retains the budget-window check, rather than checking the NO-PROMOTION phrase
twice. Apply the identical check_value change in contracts/OMN-15488.yaml lines
192-206 and
drift/dod_receipts/OMN-15488/dod-omn15488-ac4-promotion-within-budget/command.yaml
line 7, then re-run the probe to refresh the receipt.
Comment on lines
+215
to
+227
| `p = 0.070306 >= 0.05`, landing DIRECTIONAL-ONLY; the three pre-registered bands are quoted verbatim | ||
| including `NOT-SUPPORTED — D_ws <= 0.`; and the negatives carry their own section titled `first-class, | ||
| not a footnote`. Pinning the numbers rather than the verdict is the discriminator between "the pre-registered | ||
| statistic was computed and reported honestly" and "a favorable substitute statistic was reported | ||
| under the pre-registered name". Two residuals recorded honestly: the document is titled DRAFT, and | ||
| the ticket's own Deliverables section had declared the scoring document follow-on work rather than | ||
| an OMN-15488 deliverable while its AC5 requires it — it landed in #241 either way. | ||
| source: "manual" | ||
| status: "verified" | ||
| checks: | ||
| - check_type: "command" | ||
| check_value: |- | ||
| s="$(mktemp)" && gh api 'repos/jonahgabriel/steel_onslaught/contents/docs/evidence/2026-07-31-lgate2-decisive-battery-scoring.md?ref=fa2bd9d4d3d94c8bab3d3696e274ca002c284eac' --jq .content | base64 -d > "$s" && grep -qF 'D_ws = -0.003425 <= 0' "$s" && grep -qF 't = -0.570669 (one-sided Welch, direction UP)' "$s" && grep -qF 'df = 57.644039 (Welch-Satterthwaite)' "$s" && grep -qF 'D_vent = +0.019604 > 0` but `p = 0.070306 >= 0.05' "$s" && grep -qF 'NOT-SUPPORTED' "$s" && grep -qF 'first-class, not a footnote' "$s" && rm -f "$s" |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟡 Minor | ⚡ Quick win
The AC5 NOT-SUPPORTED leg is weaker than the description in both copies. The same command string appears in the contract and in its receipt. Both grep the bare token NOT-SUPPORTED rather than the pre-registered band clause that the description says is quoted verbatim.
contracts/OMN-15488.yaml#L215-L227: change the leg togrep -qF 'NOT-SUPPORTED — D_ws <= 0.' "$s".drift/dod_receipts/OMN-15488/dod-omn15488-ac5-primary-scored-as-prereg/command.yaml#L7-L7: apply the identicalcheck_valueedit and re-run the probe.
📍 Affects 2 files
contracts/OMN-15488.yaml#L215-L227(this comment)drift/dod_receipts/OMN-15488/dod-omn15488-ac5-primary-scored-as-prereg/command.yaml#L7-L7
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@contracts/OMN-15488.yaml` around lines 215 - 227, Strengthen the AC5
NOT-SUPPORTED verification by replacing the bare token grep in
contracts/OMN-15488.yaml lines 215-227 with an exact check for the
pre-registered band clause “NOT-SUPPORTED — D_ws <= 0.”; apply the identical
check_value change in
drift/dod_receipts/OMN-15488/dod-omn15488-ac5-primary-scored-as-prereg/command.yaml
line 7 and re-run the probe.
…r_number + full commit_sha) occ-preflight/eligibility failed CLOSED with reason=pr_ticket_mismatch: "no PASS receipt for one or more tickets binds to PR #5759 or one of its commit SHAs: OMN-15239, OMN-15488". Both self-bind receipts were PRESENT and PASS in the receipt_ids list, so the defect was purely in how they bound. Root cause, two fields, against the merged occ-self-bind-pr-5756 precedent: 1. neither receipt carried the pr_number field at all 2. commit_sha was the ABBREVIATED "8840584099", which matches no 40-char PR commit SHA Both fixed: pr_number: 5759, and commit_sha set to the full 8840584 (the commit that introduced the contracts, an actual commit of this PR). Same shape as the merged drift/dod_receipts/OMN-15535/occ-self-bind-pr-5756/command.yaml. verify/verify failed on the SAME single root cause -- it invokes the same eligibility resolver -- so both red checks are expected to clear together. No contract entry changed, so every contract_entry_sha256 in all 19 receipts is untouched and still hash-matches. Binding metadata only; flips no Linear status. Cites OMN-15488, OMN-15239.
jonahgabriel
added a commit
that referenced
this pull request
Jul 31, 2026
…239 (executor-compat correction) Remediation of the wave-2 OCC DoD-contract lane (OCC #5759). The four binding check_values that lane authored were rejected PRE-EXECUTION by node_dod_verify, so dod_verify returned status=failed on both tickets: OMN-15488 12 verified / 3 failed / 15 total, OMN-15239 9 verified / 1 failed / 10 total. The blocker moved from CONTRACT_MISSING to FAILED; the closeout was NOT unblocked. Root cause (not a flake, not a retry): evidence_collector._invalid_check_value_reason shape-guards a check_value before executing it by tokenizing with shlex.split(posix=True), which is not $()-aware. The embedded jq program `.state + " " + (.mergeCommit.oid // "none")` closed the outer double quote, so the guard judged the fragment `" + (.mergeCommit.oid // none)')"` as the command name, took its relative-script-path branch on the embedded slash, and returned INVALID_CHECK_VALUE_NOT_A_COMMAND without ever running the command. A false RED against a command that exits 0 in a real shell -- but a hard FAIL on the verdict. Fix: append one supersession item per defective binding, leading with a resolvable first token (`test`) and splitting the two-field query into two single-field queries so each --jq filter is a bare path expression with no quote characters. Semantics preserved exactly: BOTH state MERGED AND the exact merge SHA are asserted. Mirrors the merged OCC #5760 precedent for OMN-15535/15490/15493. Append-only, NOT an in-place edit: merged dod_evidence entries are immutable (validator_occ_append_only) and their receipts pin the entry hash. Net diff is 225 insertions / 0 deletions; no pre-existing entry or receipt is touched. Also appended: dod-omn15488-summary-correction-pr239-title, a falsifiable correction of a factual error in OMN-15488.yaml`s summary, which mislabels steel PR #239 as "Amendment 2". Its real title is "fix(OMN-15522): send programming response contract on delegation wire (r4)" and OMN-15535`s contract binds it for the stdout-tail/CLI-diagnostics fix. The SHA 4bae73ea and the "bound under OMN-15535" claim in that summary are correct and unchanged. Recorded as a check rather than PR-body prose so it cannot rot. Proof is the artifact that runs, not a bare shell surrogate. node_dod_verify was executed against these contracts: BEFORE: OMN-15488 failed (12/3/15), OMN-15239 failed (9/1/10) AFTER: OMN-15488 verified 17/17 (superseded 3), OMN-15239 verified 10/10 (superseded 1); exit 0 on both Re-run AFTER the yamlfmt reflow with identical results, and every appended receipt`s contract_entry_sha256 recomputed post-reflow and confirmed matching. Each new check_value was also RED-proven by SHA (or title) substitution: exit 1. Ticket: OMN-15488, OMN-15239
jonahgabriel
added a commit
that referenced
this pull request
Jul 31, 2026
… pr_ticket_mismatch, 3rd recurrence) occ-preflight eligibility is evaluated PER-PR: it requires a PASS receipt bound to this PR number or a head SHA. Every stamp on these tickets binds an earlier PR (#5756/#5759 merged; #5763 CLOSED UNMERGED 21:42:59Z, branch deleted), so #5767 failed occ-preflight and verify with reason=pr_ticket_mismatch. Appends occ-self-bind-pr-5767 entries + PASS receipts. Append-only: no pre-existing entry or receipt modified; entry hashes computed post-yamlfmt.
jonahgabriel
added a commit
that referenced
this pull request
Jul 31, 2026
* docs(OMN-15488): guard-compatible binding checks for OMN-15488/OMN-15239 (executor-compat correction) Remediation of the wave-2 OCC DoD-contract lane (OCC #5759). The four binding check_values that lane authored were rejected PRE-EXECUTION by node_dod_verify, so dod_verify returned status=failed on both tickets: OMN-15488 12 verified / 3 failed / 15 total, OMN-15239 9 verified / 1 failed / 10 total. The blocker moved from CONTRACT_MISSING to FAILED; the closeout was NOT unblocked. Root cause (not a flake, not a retry): evidence_collector._invalid_check_value_reason shape-guards a check_value before executing it by tokenizing with shlex.split(posix=True), which is not $()-aware. The embedded jq program `.state + " " + (.mergeCommit.oid // "none")` closed the outer double quote, so the guard judged the fragment `" + (.mergeCommit.oid // none)')"` as the command name, took its relative-script-path branch on the embedded slash, and returned INVALID_CHECK_VALUE_NOT_A_COMMAND without ever running the command. A false RED against a command that exits 0 in a real shell -- but a hard FAIL on the verdict. Fix: append one supersession item per defective binding, leading with a resolvable first token (`test`) and splitting the two-field query into two single-field queries so each --jq filter is a bare path expression with no quote characters. Semantics preserved exactly: BOTH state MERGED AND the exact merge SHA are asserted. Mirrors the merged OCC #5760 precedent for OMN-15535/15490/15493. Append-only, NOT an in-place edit: merged dod_evidence entries are immutable (validator_occ_append_only) and their receipts pin the entry hash. Net diff is 225 insertions / 0 deletions; no pre-existing entry or receipt is touched. Also appended: dod-omn15488-summary-correction-pr239-title, a falsifiable correction of a factual error in OMN-15488.yaml`s summary, which mislabels steel PR #239 as "Amendment 2". Its real title is "fix(OMN-15522): send programming response contract on delegation wire (r4)" and OMN-15535`s contract binds it for the stdout-tail/CLI-diagnostics fix. The SHA 4bae73ea and the "bound under OMN-15535" claim in that summary are correct and unchanged. Recorded as a check rather than PR-body prose so it cannot rot. Proof is the artifact that runs, not a bare shell surrogate. node_dod_verify was executed against these contracts: BEFORE: OMN-15488 failed (12/3/15), OMN-15239 failed (9/1/10) AFTER: OMN-15488 verified 17/17 (superseded 3), OMN-15239 verified 10/10 (superseded 1); exit 0 on both Re-run AFTER the yamlfmt reflow with identical results, and every appended receipt`s contract_entry_sha256 recomputed post-reflow and confirmed matching. Each new check_value was also RED-proven by SHA (or title) substitution: exit 1. Ticket: OMN-15488, OMN-15239 * docs(OMN-15488): bind OMN-15488/OMN-15239 to OCC #5763 (occ-preflight pr_ticket_mismatch repair) Adds the occ-self-bind-pr-5763 stamp both contracts were missing. Without it occ-preflight/eligibility and verify/verify fail with reason=pr_ticket_mismatch: every receipt in this PR binds OCC#5759, none binds #5763, and eligibility is evaluated per-PR not per-ticket-ever. The sibling repair PR #5760 carried the analogous occ-self-bind-pr-5760 and passed. Append-only: no pre-existing contract entry or receipt is modified. * docs(OMN-15488): bind OMN-15488/OMN-15239 to OCC #5767 (occ-preflight pr_ticket_mismatch, 3rd recurrence) occ-preflight eligibility is evaluated PER-PR: it requires a PASS receipt bound to this PR number or a head SHA. Every stamp on these tickets binds an earlier PR (#5756/#5759 merged; #5763 CLOSED UNMERGED 21:42:59Z, branch deleted), so #5767 failed occ-preflight and verify with reason=pr_ticket_mismatch. Appends occ-self-bind-pr-5767 entries + PASS receipts. Append-only: no pre-existing entry or receipt modified; entry hashes computed post-yamlfmt.
This was referenced Jul 31, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this is
Authors the two OCC DoD contracts that the Steel
dod_verifycloseout is blocked on, plus their 17 PASS receipts. Net-new files only — no existing file is touched.contracts/OMN-15488.yaml(10 evidence items)contracts/OMN-15239.yaml(7 evidence items)drift/dod_receipts/OMN-15488/*/command.yaml,drift/dod_receipts/OMN-15239/*/command.yamlScope correction vs the dispatch — 2 contracts, not 5
This lane was dispatched to author FIVE missing contracts (
OMN-15488,15239,15535,15490,15493) after a closeout run reportedCONTRACT_MISSING x5. Live readback ofonex_change_controlorigin/devat lane start found three of the five already landed in #5756 (9089ffa93):The
x5finding was accurate when measured and stale by the time this lane ran. OnlyOMN-15488andOMN-15239were genuinely absent, so this PR authors two. No duplicate contracts are created.Evidence binding
Every check pins an immutable squash-merge SHA on
jonahgabriel/steel_onslaughtmain— never a feature-branch ref, never a workflow-run conclusion.ff5df34f…,a3b0d8aa…(executing pin),fa2bd9d4…23a88d09…steel#239(4bae73ea…, Amendment 2) is already bound bycontracts/OMN-15535.yamlfrom #5756 and is deliberately not re-asserted here.The autobind trap, handled explicitly
node_dod_verify'sevidence_collectorautobinds PR references against_DEFAULT_GITHUB_ORG = "OmniNode-ai"(src/omnimarket/nodes/node_dod_verify/services/evidence_collector.py:110), which resolves nothing for a personal repo. Its fallback is_hardcoded_pr_bindings_in_value, which only accepts a(repo, pr_number)pair when both come from the same shell clause. Every check here therefore carries the PR number and--repo jonahgabriel/steel_onslaughtin one clause, e.g.Verification
All 17 checks were executed as written on the gate host (
.200), both before and after theyamlfmtpre-commit reflow — the second run is the load-bearing one, since a reflow that re-wrapped a block-scalar shell command would silently break every check. 17/17 GREEN both times.RED-proof, 17/17 flip:
OMN-15488deliverable-1 source items) pina3b0d8aa(chore(contracts): backfill ModelTicketContract for 50 done tickets [OMN-8637 wave 78] #240), whose first parente3dbd017(chore(contracts): backfill ModelTicketContract for 50 done tickets [OMN-8637 wave 75] #237) already contains deliverable 1 — it landed earlier in chore(contracts): backfill ModelTicketContract for 50 done tickets [OMN-8637 wave 73] #235. The naive first-parent oracle leaves them GREEN and proves nothing. They are RED-proven againstbe4dbe49, the first parent of chore(contracts): backfill ModelTicketContract for 50 done tickets [OMN-8637 wave 73] #235, the last commit before the overlay, the--genesisknob and the contract tests existed. All three flip RED there. This is recorded in the contract summary, not only here.Honest scope notes (carried in the contracts, not just this body)
OMN-15488AC1/AC2/AC3/AC6 attest recorded results, not re-executed gates. The battery artifacts live in a local hermetic snapshot (~/.omnibase/battery_envs/omn15488-hermetic-20260730) no hosted runner can reach, so these checks pin the independent gates agent's document at an immutable SHA. The independence claim is itself checkable — §5 of that document records three distinct agents (gates / scoring assembler / scoring recompute verifier).check_contamination_gate.py, not one whole-lane run, because that script'sload_battery_rows/check_seed_coverageinterface compares all actual seeds against a single expected block and cannot express a three-phase lane. Interface limitation, not a skipped gate — stated as such.D_ws = -0.003425,t = -0.570669,df = 57.644039) rather than a verdict word, so a favorable substitute statistic reported under the pre-registered name could not pass. The scoring document is titled DRAFT.OMN-15488deliverable-1 negative leg is scoped. No whole-file absence assertion is made for the literalmax_value=3.0, because it legitimately survives at two prose sites documenting the very defect that was fixed; the executed call site is asserted positively viamax_value=promote_cap,instead.No self-authored evidence
This lane (
fable-steel-0731wave-2) implemented none of the underlying steel work — it was built by theauthorized-wave-0730/fable-night-0730/ earlier steel lanes. Author ≠ implementer, per the no-self-authored-evidence rule.Closeout note for whoever re-runs
dod_verifyPer OMN-15369, the closeout re-run must set
ONEX_EVIDENCE_ROOTor pass--output-path— otherwise no durable receipt is written and the run proves nothing.Status
No Linear status is changed by this PR. A Done flip routes only through
dod_verifywith durable evidence. Landing belongs to Codex's merge sweep — this PR is not armed for auto-merge.Cites OMN-15488, OMN-15239.
Summary by CodeRabbit
Documentation
Tests