From 5f9b5e2043c4ec8e5cad667eaa4afbb09f1a9638 Mon Sep 17 00:00:00 2001 From: Dotta Date: Thu, 8 Oct 2026 14:05:03 -0500 Subject: [PATCH] Record installed local canary and denial fixture review proof --- ...2026-10-03-copilot-production-readiness.md | 10 ++-- .../2026-10-08-copilot-qualification-v33.json | 53 ++++++++++++++++--- 2 files changed, 52 insertions(+), 11 deletions(-) diff --git a/doc/plans/2026-10-03-copilot-production-readiness.md b/doc/plans/2026-10-03-copilot-production-readiness.md index b3ff37985ed..cfc07c4bd3d 100644 --- a/doc/plans/2026-10-03-copilot-production-readiness.md +++ b/doc/plans/2026-10-03-copilot-production-readiness.md @@ -6,7 +6,9 @@ Current release checklist, updated 2026-10-08 (America/Chicago). The frozen application source is `024e1422f9c7b43ff1b083e7d8a25e1969dc4acf`; native artifacts come from `941fbcb664a885eea713e732edc10ac133c62238`. Copilot **1.0.88**, exact **gpt-5.6-luna**, profile **v33**, and `sha256:482c4e997b4690c7e4a5375dd035560aaad69000f8f5291904d76f8c3a88284a` identify this campaign. Master baseline is `0ac194450a48a407450921a16c3ef8684dcb85ca`. Earlier results below remain historical. -The reviewed server follow-up is `50d54d78e552869cae154d88ade68063191a547e`. It preserves explicitly selected remote environments under managed-only policy and resolves forced Kubernetes before Copilot admission. All 63 focused regressions, server typecheck and server build pass. It changes neither qualified runtime/profile/image inputs nor the campaign's local and Daytona execution policies. Its ordinary installed canary is a separate remaining gate; existing live results retain their measured source revision. +The reviewed server follow-up is `50d54d78e552869cae154d88ade68063191a547e`. It preserves explicitly selected remote environments under managed-only policy and resolves forced Kubernetes before Copilot admission. All 63 focused regressions, server typecheck and server build pass. It changes neither qualified runtime/profile/image inputs nor the campaign's local and Daytona execution policies. Its ordinary installed local canary now passes separately; existing live results retain their measured source revision. The normal user path connects a personal saved token, selects the exact model, saves and reopens configuration, passes metadata Test without a prompt, completes one task, downloads the exact 30-byte artifact, and retires all owned processes without qualification overrides. + +The denial-fixture follow-up is `dc5ebe3f92d405e284b096f5add797864bb0f74e`. It retains a terminal first observed while saving the pre-Stop receipt, then checks again for an earlier Stop before dispatch. All 82 settlement regressions and Product typecheck pass. Case definitions and grading assertions are unchanged; native source, profile and image content identity are unchanged. Earlier live results retain their measured fixture/source revision. | Current gate | Result | |---|---| @@ -15,11 +17,11 @@ The reviewed server follow-up is `50d54d78e552869cae154d88ade68063191a547e`. It | Product Daytona | **0/10 current passes**; exact v33 publication approval pending; attached-command cell runs first | | Platform install/startup | Actual installed ARM64, x64 under Rosetta, and Linux x64 under Docker emulation initialize and exit cleanly; native Daytona still pending | | Authenticated metadata | Packaged runtime discovers 19 models including the exact qualification model; no prompt | -| Ordinary user path | Fresh normal package installation and local/Daytona canaries on the reviewed source remain required, without qualification overrides | +| Ordinary user path | Fresh normal package installation, saved-token setup, metadata Test, exact task completion/download and cleanup pass locally on reviewed source without qualification overrides; native Daytona remains pending | | Current image | Credential scan covered 225,852 files with zero selected credential matches. Tag `copilot-qualification-22fd2a8ae3414eee317e`; OCI index `sha256:65fc4f30c15b3d744534ce3d7240dfbf3d3ff081da6d51180bd89f2802bd6e56`; not published | -| CI / review / shipping | Runtime and regression heads passed CI and fresh 5/5 review. Setup and qualification follow-ups require current-head CI/review. Separate code-owner approval and shipped canary remain required | +| CI / review / shipping | Runtime and regression heads passed CI and fresh 5/5 review. Setup has fresh 5/5 review. Qualification requires fresh CI/review after the denial-fixture repair; pending trusted-runner selection has now completed. Separate code-owner approval and shipped canary remain required | -No automatic behavior retries occurred. Retained failures include the v32 permission-target defect, pre-provider Vite startup failures, and the private eval admission mismatch. Their original grades remain unchanged. The reconciled allowance had **$39.86 remaining** after the local campaign; provider USD stays unknown and attributed to GitHub. No budget or billing settings were reset. The current machine-readable checklist is [2026-10-08-copilot-qualification-v33.json](2026-10-08-copilot-qualification-v33.json). +No automatic behavior retries occurred. Retained failures include the v32 permission-target defect, pre-provider Vite startup failures, and the private eval admission mismatch. The first ordinary canary completed but wrote Markdown escape backslashes, producing 34 bytes instead of 30; it remains failed. One changed definition gave the same expected contents as exact hexadecimal bytes and passed. Original grades remain unchanged. The reconciled allowance has **$35.86 remaining** after these canaries; provider USD stays unknown and attributed to GitHub. No budget or billing settings were reset. The current machine-readable checklist is [2026-10-08-copilot-qualification-v33.json](2026-10-08-copilot-qualification-v33.json). ## Historical v32 release attempt diff --git a/doc/plans/2026-10-08-copilot-qualification-v33.json b/doc/plans/2026-10-08-copilot-qualification-v33.json index 4890501f328..1fc230ed850 100644 --- a/doc/plans/2026-10-08-copilot-qualification-v33.json +++ b/doc/plans/2026-10-08-copilot-qualification-v33.json @@ -1,6 +1,6 @@ { "schema": "paperclip.copilot.production-qualification/v33", - "updatedAt": "2026-10-08T18:33:47.962516+00:00", + "updatedAt": "2026-10-08T19:04:43.055832+00:00", "qualifiedApplicationSource": "024e1422f9c7b43ff1b083e7d8a25e1969dc4acf", "reviewedApplicationSource": "50d54d78e552869cae154d88ade68063191a547e", "runtimeArtifactSource": "941fbcb664a885eea713e732edc10ac133c62238", @@ -277,14 +277,26 @@ "qualifiedProfileChanged": false, "selectedLocalPolicyChanged": false, "selectedDaytonaPolicyChanged": false, - "ordinaryInstallCanary": "pending on reviewed source" + "ordinaryInstallCanary": "passed on reviewed source 50d54d78e552869cae154d88ade68063191a547e" }, "canaries": { - "ordinaryLocal": "pending", - "ordinaryDaytona": "pending" + "ordinaryLocal": "passed on reviewed source", + "ordinaryDaytona": "pending", + "ordinaryLocalEvidence": { + "sourceRevision": "50d54d78e552869cae154d88ade68063191a547e", + "definition": "Exact hexadecimal bytes; same expected file and final response", + "artifactBytes": 30, + "proofSha256": "sha256:5e826570b86129ea16055d44f99978bf2414edac4791cb7e75a72b3f14e1b43a", + "acceptedCompletionEvidence": true, + "receiptReceiverAcknowledged": true, + "finalResponseCount": 1, + "cleanupConfirmed": true, + "qualificationOverrides": [], + "previousFailureRetained": true + } }, "budget": { - "remainingConservativeAllowanceUsd": 39.86, + "remainingConservativeAllowanceUsd": 35.86, "nativeUsd": null, "billingEntity": "github", "budgetReset": false, @@ -305,15 +317,42 @@ "attempt": "v33 protocol admission attempt 1", "cause": "private eval closed allowlist lacked exact v33 profile; no provider session", "runtimeRepair": "private roster admission synchronized; all seven cases pass" + }, + { + "attempt": "v33 ordinary local canary 1", + "cause": "Provider copied Markdown underscore escapes into a 34-byte file instead of the requested 30 bytes", + "runtimeRepair": null, + "definitionFollowup": "One changed definition specified exact hexadecimal bytes; expected file and final response unchanged; follow-up passed; original failure and cleanup retained" } ], "releaseGates": [ "Exact v33 public image publication approval", "Ten exact native Daytona Product cells", - "Ordinary installed local and Daytona canaries on reviewed source", + "Ordinary installed Daytona canary on reviewed source", "Current-head CI and fresh clean review", "Separate code-owner approval, normal merge and shipped canary" ], "productionReady": false, - "shipped": false + "shipped": false, + "reviewedFixtureDelta": { + "sourceRevision": "dc5ebe3f92d405e284b096f5add797864bb0f74e", + "changedFiles": [ + "tests/runner-e2e/copilot-protection-flow.ts", + "tests/runner-e2e/copilot-protection-settlement.test.ts" + ], + "finding": 4222861745, + "focusedRegressionsPassed": 82, + "productTypecheckPassed": true, + "caseDefinitionsChanged": false, + "behaviorAssertionsWeakened": false, + "newProviderCalls": 0, + "runtimeAndImageInputsChanged": false, + "imageContentId": "22fd2a8ae3414eee317e8d9df9fa6e526e43874d590bec84081ad00ff290a472", + "contentIdentityArguments": [], + "originalLiveGradesRelabeled": false, + "sourceHashes": { + "tests/runner-e2e/copilot-protection-flow.ts": "sha256:b396a29ed3f40c398591d6687798a31234c24f94753c5c048b831a84b7fef8c7", + "tests/runner-e2e/copilot-protection-settlement.test.ts": "sha256:bdbe4bb254ddfdfd0983a164e81a6b9e33c6f79c14d653ef2446a4f6c9f61a83" + } + } }