Repository navigation
Persist Claim delivery evidence for restart-safe verification - #76718
Conversation
| receipt = { expectation, expected_digest: digest(adapterEffectFields(adapter, stored.message)) }; | ||
| } | ||
| if (!receipt || Buffer.byteLength(stored.adapter, "utf8") > 8192) return null; | ||
| return { verifier, adapter: stored.adapter, receipt }; |
| "producer": {"pool_install", "deployment", "submit", "cancel_work"}, | ||
| "dispatcher": {"pool_install", "deployment", "submit", "dispatch_next", "observe", "dispatch"}, | ||
| "worker": {"submit", "dispatch_next", "observe", "finish", "dispatch"}, | ||
| "worker": {"submit", "dispatch_next", "observe", "finish", "dispatch", "delivery_evidence"}, |
| "minimum": 0, | ||
| "type": "integer" | ||
| }, | ||
| "receipt": { |
e266243 to
0b55078
Compare
|
Category: bug | Risk: medium | Priority: medium | Score: 65/100 (impact 35 + urgency 17 + quality 13)
|
adbf649 to
5a1fe7a
Compare
|
✅ PR Code Quality Reviewer completed the code quality review. PR Code Quality ReviewerOutcome: The agent completed and reported one or more actions. The no-op output is a completion signal, not a claim that no action was taken. Agent messages are omitted from this comment. The category uses structured run status, not interpretation of agent prose.
|
|
🧠 Matt Pocock Skills Reviewer has completed the skills-based review. ✅
|
|
✂️ Ponytail Reviewer has started processing this pull request |
|
✅ Design Decision Gate 🏗️ completed the design decision gate check. See the comment below for the result and any generated ADR draft.
|
|
✅ Test Quality Sentinel completed test quality analysis. Test Quality SentinelOutcome: The agent completed and reported no action. A no-op output alone does not verify that all required work was completed or that publication was possible. Agent messages are omitted from this comment. The category uses structured run status, not interpretation of agent prose.
|
|
✅ PR Specification Reviewer completed successfully! PR Specification ReviewerOutcome: The agent completed and reported no action. A no-op output alone does not verify that all required work was completed or that publication was possible. Agent messages are omitted from this comment. The category uses structured run status, not interpretation of agent prose.
|
Comment MemoryPeek at saved memory (pr-code-quality-reviewer)Note This comment is managed by comment memory.Expand the saved memory block to view or edit the persistent context for this thread.
|
There was a problem hiding this comment.
Found one blocking correctness issue in durable code-scanning evidence: the validator checks receipt.sha256 although receipts store finding_digest, so restart-safe evidence is never published for code-scanning outputs. The inline comment identifies the fix. Specialist pass was unavailable because its file-read tool call was denied; this review is based on the pre-fetched diff and existing comments.
🔎 Code quality review by PR Code Quality Reviewer · copilot · gpt50 · ~3 AIC · ⌖ ~2 AIC · ⊞ 1.6K
Comment /review to run again
| receiptMatchesClaim(receipt, claim) && | ||
| ["repository", "revision", "ref", "driver", "category", "sarif_id"].every(key => typeof receipt[key] === "string" && receipt[key].length > 0) && | ||
| /^[A-Za-z0-9_-]+\/[A-Za-z0-9._-]+$/.test(receipt.repository) && | ||
| /^[a-f0-9]{64}$/.test(receipt.finding_digest) && |
There was a problem hiding this comment.
main() stores the receipt field as finding_digest (line 95), but this validator checks receipt.sha256, which is never populated. Consequently every code-scanning receipt fails validCodeScanningReceipt, durableCodeScanningReceipt() returns null, and collectDeliveryEvidence() publishes delivery_evidence_unsupported:create_code_scanning_alert; maintenance can never perform restart-safe verification for this output type. Validate finding_digest here (and retain the digest check below).
There was a problem hiding this comment.
The current branch already validates receipt.finding_digest, matching the stored field. The existing specialized-delivery test covers the 64-hex digest and durable verification, so no additional code change was needed.
✅ Design Decision Gate — ADR VerifiedADR reviewed: ADR-76718: Retain durable Claim delivery evidence — implementation aligns with the stated decision. Great work! 🏗️ 📋 Verification SummaryThe ADR Decision commitments are implemented as written:
Divergences: none found. Notes for maintainers (not divergences):
Verification was limited to static inspection of the PR diff and the ADR. Tests were not run.
|
There was a problem hiding this comment.
Skills-Based Review 🧠
Applied /tdd and /grill-with-docs (selected by triage as new_feature). Posting observations, not blocking changes. The main open question is the controls_digest skip.
📋 Key Themes & Highlights
Key Themes
- Restart-verification gap (
/tdd): an omittedcontrols_digestskips the control-inventory check on recovery. Inline comment onwork_queue_delivery_recovery.cjs:43. - Trust boundary not documented (
/grill-with-docs): adapter config is persisted into the ledger. Low risk because it is closed and trusted, but the ADR should say so. - Silent publish degradation (
/diagnosing-bugs): a failed evidence publish only logs a warning. The fallback toDeliveryFailureis safe, but a regression would be easy to miss. - Receipt schema is open (not inline):
DeliveryEvidenceOperation.jsonusesadditionalProperties: {}. Consider closed per-verifier receipt schemas. Verifiers re-validate receipts, so this is defense-in-depth, not an exploit.
Positive Highlights
- ✅ Scope binding is enforced via
scopedClaim(dispatch and run attempt), and the per-kind permission allowlist is correct. - ✅ Size bounds (32 KiB, 128 outputs, 256 effects, 8 KiB adapter) are enforced both in Go and in JS replay.
- ✅ Negative-scope and immutability tests exist (
delivery_evidence_test.go,work_queue_delivery_evidence.test.cjs). - ✅
go test ./pkg/workqueue/...passes. - ✅ The ADR captures the decision and its trade-offs.
Limits of This Review
- The JS vitest suite was not run.
npm cifailed on the sandbox TLS proxy, so the JS side was reviewed by reading only. - The base-commit diff was unavailable (partial clone), so the review used the final file state and the capped 3000-line patch.
- The
declared_builtinrouting throughverifyDeclaredOutputwas checked by reading only.
@copilot please address the review comments above.
🧠 Reviewed using Matt Pocock's skills by Matt Pocock Skills Reviewer · copilot · routed: haiku55 xhigh · ~12 AIC · ⌖ ~2 AIC · ⊞ 6.6K
Comment /matt to run again
| if (evidence.unsupported) core.info(`Claim ${handle}: durable delivery evidence not published (${evidence.unsupported})`); | ||
| else await publishDeliveryEvidence({ assignment: scope.assignment, claim_handle: handle, ...evidence, github: options.queueClient || github, context: verificationContext, core }); | ||
| } catch (error) { | ||
| core.warning(`Claim ${handle}: durable delivery evidence publication failed: ${getErrorMessage(error)}`); |
There was a problem hiding this comment.
[/diagnosing-bugs] A failed evidence publish only logs a warning. Recovery then degrades to a DeliveryFailure, which is a safe fallback, but nothing makes a publish regression visible.
💡 Suggested change
Surface publish failures at a higher severity, or count them in run output, so a restart-safety regression shows up in CI. Add a test that asserts the warning path leaves the claim in DeliveryFailure rather than silently succeeding.
@copilot please address this.
There was a problem hiding this comment.
Updated in d39c8e4: evidence-publication failures now emit an error annotation, and a restart test confirms no Result is published and the claim eventually settles as DeliveryFailure.
| receipt = { expectation, expected_digest: digest(adapterEffectFields(adapter, stored.message)) }; | ||
| } | ||
| if (!receipt || Buffer.byteLength(stored.adapter, "utf8") > 8192) return null; | ||
| return { verifier, adapter: stored.adapter, receipt }; |
There was a problem hiding this comment.
[/grill-with-docs] This persists the adapter config into the durable ledger. The risk is low because validateAdapter limits it to closed keys and a fixed target-repo, but that trust boundary is not stated.
💡 Suggested change
State in the ADR (docs/adr/76718-durable-claim-delivery-evidence.md) that adapter config is trusted, bounded, and persisted by design. Then any future adapter key that could carry a secret gets reviewed against that rule.
@copilot please address this.
There was a problem hiding this comment.
Documented in d39c8e4 that adapter objects are persisted only after closed-key validation and the 8 KiB bound; future fields that could carry credentials or personal data require review.
| requireControlInventory: true, | ||
| readControlInventory: async () => { | ||
| const inventory = await readDeliveryControlInventory({ ...options, assignment, claim_handle: member.handle }); | ||
| if (evidence.controls_digest !== undefined && inventory.controls_digest !== evidence.controls_digest) throw new Error("work_queue_control_inventory_changed"); |
There was a problem hiding this comment.
[/tdd] When controls_digest is absent, this check is skipped, so a restarted run can accept evidence without verifying the control inventory.
💡 Suggested fix
Require controls_digest whenever the evidence carries queue_control verification. If omission is intentional, document why it is safe and add a test that pins the accepted-without-digest behavior. The existing Go test at delivery_evidence_test.go:89-91 only covers the acceptance path.
@copilot please address this.
There was a problem hiding this comment.
Updated in d39c8e4: queue-control evidence without controls_digest now fails closed before inventory readback, with restart regression coverage.
There was a problem hiding this comment.
🟡 Changes recommended
Critical recovery-verification gaps can promote rejected or failed deliveries to successful Results and incorrectly terminate valid deliveries as DeliveryFailures.
9 open issues
Grant maintenance credentials the required native read permissions · New Treat missing evidence as verifier unavailability · New Suppress evidence publication for disputed Claims · New Register artifact readback proofs with resource authority · New Check outcomes and Claims before the no-write shortcut · New Bind durable control receipts to the verified request parameters · New Pass restored effects into output verification · New Add reader-before-writer upgrade guidance for DeliveryEvidence · New Count only Results newly published by durable recovery · New
What changed in this PR
Adds durable delivery evidence to gh-aw’s work queue so maintenance can verify completed Claims after worker process exit without repeating writes.
Changes:
- Defines bounded, worker-authenticated evidence with matching schemas, replay, and checkpoint support.
- Adds receipt serialization, maintenance readback, and restart-focused verification coverage.
- Documents the recovery contract and Go 1.27.4 security baseline.
| File | Description |
|---|---|
specs/work-queue/transactions.tsp |
Defines evidence operations and receipt contracts. |
scripts/check-go-toolchain_test.sh |
Extends accepted Go patch-version coverage. |
pkg/workqueue/types.go |
Adds evidence to projection state. |
pkg/workqueue/selection.go |
Preserves evidence when cloning projections. |
pkg/workqueue/schema/RequestParameters.json |
Adds evidence request parameters. |
pkg/workqueue/schema/QueueRequest.json |
Allows evidence requests. |
pkg/workqueue/schema/QueueOperation.json |
Includes the evidence operation. |
pkg/workqueue/schema/QueueCommit.json |
Includes evidence in commit validation. |
pkg/workqueue/schema/OperationsParameters.json |
Includes evidence operation parameters. |
pkg/workqueue/schema/DeliveryEvidenceWriteOutput.json |
Defines write-receipt validation. |
pkg/workqueue/schema/DeliveryEvidenceOutput.json |
Defines the output receipt union. |
pkg/workqueue/schema/DeliveryEvidenceOperation.json |
Defines evidence identity and bounds. |
pkg/workqueue/schema/DeliveryEvidenceNoWriteOutput.json |
Defines no-write output entries. |
pkg/workqueue/schema/DeliveryEvidenceExpected.json |
Defines effect readback expectations. |
pkg/workqueue/schema/DeliveryEvidenceEffect.json |
Defines recorded effect metadata. |
pkg/workqueue/request.go |
Permits worker evidence requests. |
pkg/workqueue/replay.go |
Applies evidence and clears terminal entries. |
pkg/workqueue/policy.go |
Initializes evidence and enforces worker-only publication. |
pkg/workqueue/policy_permitted_test.go |
Tests evidence operation permissions. |
pkg/workqueue/delivery_evidence.go |
Validates evidence scope, immutability, and bounds. |
pkg/workqueue/delivery_evidence_parity_test.go |
Tests cross-runtime replay and restart parity. |
pkg/workqueue/checkpoint.go |
Retains and validates pending evidence. |
docs/src/content/docs/specs/work-queue-specification.md |
Specifies durable verification and expiry behavior. |
docs/src/content/docs/reference/work-queue.md |
Documents maintenance recovery behavior. |
docs/adr/76718-durable-claim-delivery-evidence.md |
Records design decisions and tradeoffs. |
CHANGELOG.md |
Documents the patched Go baseline. |
actions/setup/js/work_queue_upload_assets.cjs |
Adds serializable asset receipts and readback. |
actions/setup/js/work_queue_specialized_delivery.test.cjs |
Tests specialized receipt serialization and readback. |
actions/setup/js/work_queue_rest_adapter.test.cjs |
Tests durable REST receipts and bounds. |
actions/setup/js/work_queue_rest_adapter.cjs |
Adds receipt-explicit REST verification. |
actions/setup/js/work_queue_replay.test.cjs |
Tests evidence replay and checkpoint lifecycle. |
actions/setup/js/work_queue_replay.cjs |
Adds JavaScript evidence validation and retention. |
actions/setup/js/work_queue_reconciler.cjs |
Supports assignment-specific verifier factories. |
actions/setup/js/work_queue_policy.cjs |
Allows worker evidence publication. |
actions/setup/js/work_queue_native_creation_delivery.test.cjs |
Tests serialized native-creation receipts. |
actions/setup/js/work_queue_native_creation_delivery.cjs |
Adds digest-based discussion and PR receipts. |
actions/setup/js/work_queue_graphql_adapter.cjs |
Adds durable GraphQL receipt verification. |
actions/setup/js/work_queue_graphql_adapter_checks.cjs |
Tests GraphQL restart readback and bounds. |
actions/setup/js/work_queue_git_tree_adapter.cjs |
Adds digest-based durable Git receipts. |
actions/setup/js/work_queue_git_tree_adapter_checks.cjs |
Tests Git receipt recovery and mutation rejection. |
actions/setup/js/work_queue_delivery_recovery.cjs |
Assembles read-only maintenance verification. |
actions/setup/js/work_queue_delivery_evidence.test.cjs |
Tests evidence collection and publication. |
actions/setup/js/work_queue_delivery_evidence.test_helpers.cjs |
Provides worker/restart maintenance fixtures. |
actions/setup/js/work_queue_delivery_evidence.cjs |
Collects, publishes, and routes durable receipts. |
actions/setup/js/work_queue_delivery_evidence_restart.test.cjs |
Tests restart settlement without repeated writes. |
actions/setup/js/work_queue_delivery_evidence_native_restart.test.cjs |
Tests discussion and PR recovery after restart. |
actions/setup/js/work_queue_control_receipts.test.cjs |
Tests trusted control-inventory verification. |
actions/setup/js/work_queue_control_receipts.cjs |
Adds committed queue-control readback. |
actions/setup/js/work_queue_code_scanning.cjs |
Adds digest-based SARIF delivery receipts. |
actions/setup/js/work_queue_code_coverage.cjs |
Adds durable coverage receipts and validation. |
actions/setup/js/work_queue_claim_scope.test.cjs |
Tests explicit durable receipt rebinding. |
actions/setup/js/work_queue_claim_scope.cjs |
Shares receipt selection and serialization helpers. |
actions/setup/js/work_queue_claim_adapters.test.cjs |
Tests durable declared-builtin adapters. |
actions/setup/js/work_queue_claim_adapters.cjs |
Adds durable adapter export and verifier construction. |
actions/setup/js/upload_artifact.cjs |
Adds artifact receipt serialization and readback. |
actions/setup/js/upload_artifact_work_queue.test.cjs |
Tests artifact receipts after module reload. |
actions/setup/js/safe_output_handler_manager.cjs |
Publishes evidence after worker handler passes. |
actions/setup/js/reconcile_work_queue_run.cjs |
Enables maintenance recovery and reporting. |
actions/setup/js/finish_work_queue_claim.cjs |
Adds reconciler authorization and final readback handling. |
.changeset/compiler-transport-trust.md |
Updates the documented Go baseline. |
Review effort: Max (auto)
| configured.core?.warning("Work queue obsolete-work reconciliation failed; core recovery will continue. Inspect safe API diagnostics and the checked ledger before retrying maintenance."); | ||
| } | ||
| const result = await reconcileQueue({ ...configured, requestCancellation: false, expireUnverified: true }); | ||
| const result = await reconcileQueue({ ...configured, requestCancellation: false, expireUnverified: true, verifyEffectsFor: options.verifyEffectsFor || (assignment => createDurableDeliveryVerifier({ ...configured, assignment })) }); |
| configured.core?.warning("Work queue obsolete-work reconciliation failed; core recovery will continue. Inspect safe API diagnostics and the checked ledger before retrying maintenance."); | ||
| } | ||
| const result = await reconcileQueue({ ...configured, requestCancellation: false, expireUnverified: true }); | ||
| const result = await reconcileQueue({ ...configured, requestCancellation: false, expireUnverified: true, verifyEffectsFor: options.verifyEffectsFor || (assignment => createDurableDeliveryVerifier({ ...configured, assignment })) }); |
| if (options.registerVerification !== undefined && !isStagedMode()) { | ||
| try { | ||
| const evidence = collectDeliveryEvidence({ assignment: scope.assignment, claim_handle: handle, messages, results, effects: options.effects || [], controls, config: loadConfig(), context: verificationContext }); | ||
| if (evidence.unsupported) core.info(`Claim ${handle}: durable delivery evidence not published (${evidence.unsupported})`); | ||
| else await publishDeliveryEvidence({ assignment: scope.assignment, claim_handle: handle, ...evidence, github: options.queueClient || github, context: verificationContext, core }); |
| async function verifyArtifactReceipt({ claim, github }, receipt) { | ||
| if (!validArtifactReceipt(receipt, claim)) return { verified: false }; |
| if (NO_WRITE_TYPES.has(type)) return { entry: { index, type, verifier: "none" } }; | ||
| const outcome = results.find(result => result.messageIndex === index); | ||
| // Omitting a failed optional output would turn an incomplete pass into a | ||
| // complete synthetic inventory after restart. Never publish that subset. | ||
| if (!outcome || !outcome.success || outcome.delegated || outcome.deferred || outcome.skipped || outcome.cancelled) return { unsupported: `delivery_evidence_receipt_missing:${type}` }; | ||
| if (outcome.claim_handle && outcome.claim_handle !== claim.handle) throw new Error("work_queue_claim_scope_invalid"); |
| if (message.claim_handle !== claim.handle || !["work_queue_submit", "work_queue_dispatch_next"].includes(message.type)) throw new Error("work_queue_control_receipt_scope_invalid"); | ||
| // request_id is filled from authenticated worker authority at publication, | ||
| // not the unauthenticated Actions context supplied to the handler manager. | ||
| outputs.push({ index: messages.length + offset, type: message.type, verifier: "queue_control", receipt: { intent_id: message.intent_id } }); |
| return inventory; | ||
| }, | ||
| verifyDeclaredOutput: createDurableDeclaredVerifier(evidence.outputs, effects), | ||
| verifyOutput: verifyEvidenceEntry, |
| ], | ||
| Claim: [["kind", "work_id", "claim_id", "dispatch_id", "handle", "observations"], []], | ||
| Completion: [["kind", "work_id", "claim_id", "dispatch_id", "claim_handle", "run_id", "run_attempt"], ["keep_open"]], | ||
| DeliveryEvidence: [["kind", "work_id", "claim_id", "dispatch_id", "claim_handle", "completion_id", "run_id", "run_attempt", "outputs", "effects"], ["controls_digest"]], |
| reconciled: result.reconciled, | ||
| released, | ||
| unresolved, | ||
| durable_verified: result.results.filter(entry => Object.values(entry.delivery?.claims || {}).some(claim => claim.state === "result")).length, |
|
🍳 A little prep to help this PR along! Copilot assignment requested for the following outstanding work:
Keeping the changes focused; human review and CI get the final taste test. No merging or CI triggering. Structured data{
"completed_task": 0,
"head": "5a1fe7a7b7f854b1fffd80150d215addbe1b1b70",
"kind": "requested",
"schema_version": "1",
"state": "7a7d724d577d90e1f70efd22533ab8540c0b7833bbc5c62a431c455529508934",
"work": [
"14d1bff15b137e43e6d04c254a5005af5064357e69aa4d379c56d9ebd9780c3a",
"51e2287c3d1ee41b7e13769862add61dcd292a623cf943abc5968c984e246bcf",
"ae9e6da09f42352b7fb62950a192aa6069720befae086ea83c724bcafabad2ab",
"ee1e3f66cbe756e4582e3ef4d93cf09cfe41bc8042de8dfe99033b074949ee7c",
"f2a1d74d275a90c1cd44b26a5be86d15ba58b0cfefb0980afac2aa391410b7d4"
],
"workflow": "pr-sous-chef"
}
|
Co-authored-by: gh-aw-bot <259018956+gh-aw-bot@users.noreply.github.com>
d39c8e4 to
9e1bb02
Compare
|
🎉 This pull request is included in a new release. Release: |


The reported ledger contained 13 Work items that expired as
DeliveryFailure(verification_exhausted)because delivery receipts lived in process-local WeakMaps and maintenance had no verifier. This change publishes bounded, authenticatedDeliveryEvidenceafter the worker handler pass so maintenance can verify delivery by fresh native readback after process exit, without repeating writes or rewriting historical terminal decisions.Review map
flowchart LR W["Worker safe_outputs<br/>handler pass"] -- "B3 serializable receipts" --> P["collect + publish<br/>(B4)"] P -- "B1 DeliveryEvidence op" --> L[("checked ledger")] L -. "B2 checkpoint keeps it<br/>while barrier pending" .-> L M["Scheduled maintenance<br/>(B5)"] -- "reads evidence" --> L M -- "fresh native readback" --> G["GitHub resources"] M -- "Result or DeliveryFailure" --> Lactions/setup/js/work_queue_policy.cjsactions/setup/js/work_queue_replay.cjspkg/workqueue/delivery_evidence.gopkg/workqueue/policy.gopkg/workqueue/replay.gopkg/workqueue/request.gopkg/workqueue/schema/DeliveryEvidenceEffect.jsonpkg/workqueue/schema/DeliveryEvidenceExpected.jsonpkg/workqueue/schema/DeliveryEvidenceNoWriteOutput.jsonpkg/workqueue/schema/DeliveryEvidenceOperation.jsonpkg/workqueue/schema/DeliveryEvidenceOutput.jsonpkg/workqueue/schema/DeliveryEvidenceWriteOutput.jsonpkg/workqueue/schema/OperationsParameters.jsonpkg/workqueue/schema/QueueCommit.jsonpkg/workqueue/schema/QueueOperation.jsonpkg/workqueue/schema/QueueRequest.jsonpkg/workqueue/schema/RequestParameters.jsonpkg/workqueue/types.gospecs/work-queue/transactions.tspactions/setup/js/work_queue_replay.test.cjspkg/workqueue/delivery_evidence_test.gopkg/workqueue/policy_permitted_test.gopkg/workqueue/checkpoint.gopkg/workqueue/selection.gopkg/workqueue/delivery_evidence_parity_test.goactions/setup/js/upload_artifact.cjsactions/setup/js/work_queue_claim_adapters.cjsactions/setup/js/work_queue_claim_scope.cjsactions/setup/js/work_queue_code_coverage.cjsactions/setup/js/work_queue_code_scanning.cjsactions/setup/js/work_queue_control_receipts.cjsactions/setup/js/work_queue_delivery.cjsactions/setup/js/work_queue_git_tree_adapter.cjsactions/setup/js/work_queue_graphql_adapter.cjsactions/setup/js/work_queue_native_creation_delivery.cjsactions/setup/js/work_queue_rest_adapter.cjsactions/setup/js/work_queue_upload_assets.cjsactions/setup/js/upload_artifact_work_queue.test.cjsactions/setup/js/work_queue_claim_adapters.test.cjsactions/setup/js/work_queue_claim_scope.test.cjsactions/setup/js/work_queue_control_receipts.test.cjsactions/setup/js/work_queue_delivery.test.cjsactions/setup/js/work_queue_git_tree_adapter_checks.cjsactions/setup/js/work_queue_graphql_adapter_checks.cjsactions/setup/js/work_queue_native_creation_delivery.test.cjsactions/setup/js/work_queue_rest_adapter.test.cjsactions/setup/js/work_queue_specialized_delivery.test.cjsactions/setup/js/safe_output_handler_manager.cjsactions/setup/js/work_queue_delivery_evidence.cjsactions/setup/js/work_queue_delivery_evidence.test.cjsactions/setup/js/finish_work_queue_claim.cjsactions/setup/js/reconcile_work_queue_run.cjsactions/setup/js/work_queue_delivery_recovery.cjsactions/setup/js/work_queue_reconciler.cjsactions/setup/js/work_queue_delivery_evidence.test_helpers.cjsactions/setup/js/work_queue_delivery_evidence_adapters_restart.test.cjsactions/setup/js/work_queue_delivery_evidence_native_restart.test.cjsactions/setup/js/work_queue_delivery_evidence_restart.test.cjsdocs/adr/76718-durable-claim-delivery-evidence.mddocs/src/content/docs/reference/work-queue.mddocs/src/content/docs/specs/work-queue-specification.md| CI — Patched Go 1.27.4 baseline for GO-2026-6859 | adbf649e16 |
.changeset/compiler-transport-trust.md.devcontainer/devcontainer.json.devcontainer/upgrade-tools.sh.github/workflows/daily-code-debt-aider.lock.yml.github/workflows/daily-code-debt-aider.md.github/workflows/daily-go-test-stubs-aider.lock.yml.github/workflows/daily-go-test-stubs-aider.md.github/workflows/go.mod.github/workflows/smoke-atomic.lock.yml.github/workflows/smoke-atomic.md.github/workflows/smoke-claude.lock.yml.github/workflows/smoke-claude.md.github/workflows/smoke-codex.lock.yml.github/workflows/smoke-codex.md.github/workflows/smoke-copilot-aoai-apikey.lock.yml.github/workflows/smoke-copilot-aoai-apikey.md.github/workflows/smoke-copilot-aoai-entra.lock.yml.github/workflows/smoke-copilot-aoai-entra.md.github/workflows/smoke-copilot-arm.lock.yml.github/workflows/smoke-copilot-arm.md.github/workflows/smoke-copilot.lock.yml.github/workflows/smoke-copilot.md.github/workflows/smoke-cursor.lock.yml.github/workflows/smoke-cursor.md.github/workflows/smoke-factory.lock.yml.github/workflows/smoke-factory.md.github/workflows/smoke-goose.lock.yml.github/workflows/smoke-goose.md.github/workflows/smoke-kiro.lock.yml.github/workflows/smoke-kiro.md.github/workflows/smoke-test-tools.lock.yml.github/workflows/smoke-test-tools.md.golangci.ymlCHANGELOG.mdgo.modpkg/constants/version_constants.goscripts/check-go-toolchain.shspecs/security-architecture-spec.md|pkg/workflow/aider_discovery_test.gopkg/workflow/runtime_go_baseline_test.go(removed source-only assertions)pkg/workflow/runtime_setup_test.goscripts/check-go-toolchain_test.sh|B6 docs: specification §7.6 and §7.15, work-queue reference, ADR-76718.
Review notes
nolint:largefuncrather than a lint-driven file split.Verification
Refactored and rebuilt tips: 25 JS files / 769 tests passed; full Go workqueue package passed, including manager-restart parity and schema reproducibility. All six DeliveryEvidence commits passed
go vet ./pkg/workqueueand loading their changed production CJS modules. The six-commit rebuilt tree matched the refactored snapshot exactly. The CI repair adds 42 files; all 103 changed files are attributed once above.Restart proof: worker evidence survives cleared CJS modules; maintenance returns
resultwith one total write. Tampered readback yields one DeliveryFailure and zero Results. The four evidence suites pass 26 tests; the moved receipt-rebinding test passes in claim-scope. Old and regenerated evidence schemas agree on 123 boundary cases.CI repair: Go 1.27.4 fixes GO-2026-6859 / CVE-2026-97054. The exact failing scan now reports 0 reachable vulnerabilities. Release-floor tests reject 1.27.3 and accept 1.27.4; runtime selection/root tests and all 365 workflow compilations pass. All 14 affected generated locks select the patched version. Source-only/default-version echo assertions were removed rather than re-pinned; executed release-floor and runtime behavior checks remain.
make agent-report-progress-no-testpasses. GitHub CI onadbf649e16completed with 60 successful checks, 15 skipped checks, and no failures or pending checks, including the vulnerability scan, all eight JS shards, all five Go test shards, macOS/Windows builds, WASM, and Go canary.Approved upstream-only gate exception:
make agent-report-progresspassed Go/custom lint, JS lint (0 errors, 228 warnings), 5 impacted Go tests, 376 impacted JS tests, and compilation of all 365 workflows. It failed only on description drift in unchangedponytail-reviewer.lock.yml,pr-code-quality-reviewer.lock.yml, andpr-specification-reviewer.lock.yml, introduced by main commitf33f7dccc7(#76610). Those unrelated locks remain outside this PR.Structure checks: exactly the manager and recovery imports; zero obsolete symbols; three pre-existing checkpoint test files, no split production files. Reproduce with:
Publication checks (seven commits: six DeliveryEvidence behaviors plus the CI repair):
ADR-76718