From f2f3bd6a01e7f549b8706d9b2a362e037d5c6827 Mon Sep 17 00:00:00 2001 From: Claude Date: Mon, 9 Nov 2026 17:33:51 +0000 Subject: [PATCH 1/3] docs: route entry pages to CBZ-0 governance and board #738 CLAUDE.md, CONTRIBUTING.md and docs/WORKSPACE.md still called OP-WORKFLOW-20260930 the "current workflow" and #619 the active board. AGENTS.md (fixed in #739) and GOVERNANCE.md make CBZ-0 governance the current authority and #738 the active board; OP-WORKFLOW is historical where it conflicts. Routing text only: the prior workflow and older boards stay linked as history, and no rule, status or proof text changes. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01Fij9F1cYrAUDW7PhWQze1G --- CLAUDE.md | 7 +++++-- CONTRIBUTING.md | 24 +++++++++++++----------- docs/WORKSPACE.md | 16 ++++++++++------ 3 files changed, 28 insertions(+), 19 deletions(-) diff --git a/CLAUDE.md b/CLAUDE.md index 30fc7c2d1..9a6935ab9 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -1,7 +1,10 @@ # Claude entry -Use [AGENTS.md](AGENTS.md) and the -[current workflow](governance/OP-WORKFLOW-20260930.md). +Use [AGENTS.md](AGENTS.md), [CBZ-0 governance](GOVERNANCE.md) (the current +authority) and the active coordination board, +[main #738](https://github.com/d6g8k5htny-coder/main/issues/738). The +[prior workflow](governance/OP-WORKFLOW-20260930.md) is historical where it +conflicts. The same owner delegation and scope coordination apply to every participating model; there is no separate Claude permission queue. Use [Workspace and tools](docs/WORKSPACE.md) for execution paths and the diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index f2a412eb4..c521d2ed0 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -6,24 +6,26 @@ participants and may be improved by them. ## Before you start -1. Read [AGENTS.md](AGENTS.md) for the cross-model entry point and the - [current workflow](governance/OP-WORKFLOW-20260930.md) for claims, - dispositions, verification tiers and review rules; [CLAUDE.md](CLAUDE.md) - only points there. +1. Read [AGENTS.md](AGENTS.md) for the cross-model entry point and + [CBZ-0 governance](GOVERNANCE.md) for the current coordination, + verification and review rules. The + [prior workflow](governance/OP-WORKFLOW-20260930.md) is historical where it + conflicts; [CLAUDE.md](CLAUDE.md) only points there. 2. Read the [research guide](docs/RESEARCH_INDEX.md) for source-linked research context and each result's stated scope. For current work, consult [main #666](https://github.com/d6g8k5htny-coder/main/issues/666) (continued from [#275](https://github.com/d6g8k5htny-coder/main/issues/275)) for architecture decisions, the - [active coordination board #619](https://github.com/d6g8k5htny-coder/main/issues/619) - (continued from [#539](https://github.com/d6g8k5htny-coder/main/issues/539), which continued from - [#475](https://github.com/d6g8k5htny-coder/main/issues/475) and [#419](https://github.com/d6g8k5htny-coder/main/issues/419)) + [active coordination board #738](https://github.com/d6g8k5htny-coder/main/issues/738) + (continued from [#619](https://github.com/d6g8k5htny-coder/main/issues/619), which continued from + [#539](https://github.com/d6g8k5htny-coder/main/issues/539), [#475](https://github.com/d6g8k5htny-coder/main/issues/475) and [#419](https://github.com/d6g8k5htny-coder/main/issues/419)) for current coordination and bounded-task discussion, and the relevant task-local issue or PR for its latest claim, disposition and evidence. 3. **Declare the work before you begin it.** Say what you intend to change, with a timestamp, where other contributors will see it — the task and claim queue on the - [active coordination board, main #619](https://github.com/d6g8k5htny-coder/main/issues/619), or a comment on + [active coordination board, main #738](https://github.com/d6g8k5htny-coder/main/issues/738), or a comment on the relevant issue or pull request (general coordination: - [active coordination board, main #619](https://github.com/d6g8k5htny-coder/main/issues/619); see the + [active coordination board, main #738](https://github.com/d6g8k5htny-coder/main/issues/738); see the + [prior #619 board](https://github.com/d6g8k5htny-coder/main/issues/619), the [prior #539 board](https://github.com/d6g8k5htny-coder/main/issues/539), the [prior #475 board](https://github.com/d6g8k5htny-coder/main/issues/475), the [prior #419 board](https://github.com/d6g8k5htny-coder/main/issues/419), the @@ -37,7 +39,7 @@ participants and may be improved by them. Use [public tasks](https://github.com/d6g8k5htny-coder/main/issues?q=is%3Aissue+is%3Aopen+label%3Aeng-only) (none open on 7 October 2026), the -[current task and claim queue, main #619](https://github.com/d6g8k5htny-coder/main/issues/619) +[current task and claim queue, main #738](https://github.com/d6g8k5htny-coder/main/issues/738) or [propose a bounded task](https://github.com/d6g8k5htny-coder/main/issues/new?template=task.yml). The starter types are `replay`, `chart`, `review-comment`, `catalog-stub`, and `docs`. Cite the existing key, public path, full commit and digest from @@ -167,7 +169,7 @@ back regardless of how good the result looks. Read the [formal-verification guide](docs/FORMAL_VERIFICATION.md) first and claim work in the relevant formal pull request; use -[active coordination board, main #619](https://github.com/d6g8k5htny-coder/main/issues/619) to find current +[active coordination board, main #738](https://github.com/d6g8k5htny-coder/main/issues/738) to find current tasks and link that claim when coordination requires it. [#95](https://github.com/d6g8k5htny-coder/main/issues/95) is closed formal-rollout history. The diff --git a/docs/WORKSPACE.md b/docs/WORKSPACE.md index 29ca4cd4f..deff25683 100644 --- a/docs/WORKSPACE.md +++ b/docs/WORKSPACE.md @@ -4,8 +4,10 @@ For an introduction to the mathematics, begin at the [public research home](site/index.html). The [source workspace](site/workspace.html) exposes the existing technical views. This guide is for contributors who need repository locations and execution -commands. Use the [current workflow](../governance/OP-WORKFLOW-20260930.md) for -ownership, review reuse, testing and delivery. +commands. Use [CBZ-0 governance](../GOVERNANCE.md) for the current ownership, +review, testing and delivery rules; the +[prior workflow](../governance/OP-WORKFLOW-20260930.md) is historical where it +conflicts. ## Current working locations @@ -59,7 +61,8 @@ permission wording in an older mirror does not revive revoked owner restrictions ## Where coordination happens New general questions, obstacles and handoffs go on the -[active Agent Message Board, main #619](https://github.com/d6g8k5htny-coder/main/issues/619). +[active Agent Message Board, main #738](https://github.com/d6g8k5htny-coder/main/issues/738). +[main #619](https://github.com/d6g8k5htny-coder/main/issues/619), [main #539](https://github.com/d6g8k5htny-coder/main/issues/539), [main #475](https://github.com/d6g8k5htny-coder/main/issues/475), [main #419](https://github.com/d6g8k5htny-coder/main/issues/419), @@ -69,7 +72,7 @@ boards; start with the [working guidance](https://github.com/d6g8k5htny-coder/main/issues/229#issuecomment-6038554880). GitHub stopped accepting comments on #229 at 2,500. Deduplicated tasks, current claims and handoffs go to the -[active coordination board, main #619](https://github.com/d6g8k5htny-coder/main/issues/619) and the relevant +[active coordination board, main #738](https://github.com/d6g8k5htny-coder/main/issues/738) and the relevant issue or pull request; architecture discussion continues on [main #666](https://github.com/d6g8k5htny-coder/main/issues/666) (continued from [#275](https://github.com/d6g8k5htny-coder/main/issues/275)). Most proof review happens in [Math- pull requests](https://github.com/d6g8k5htny-coder/Math-/pulls). Dated queue @@ -280,7 +283,8 @@ anchors agree. That is formal evidence for exactly those targets, not mathematic acceptance; the [guide](FORMAL_VERIFICATION.md) and the [rollout record](FORMAL_VERIFICATION_ROLLOUT_20260927.md) explain the boundary. -[Current authority](../governance/OP-AUTONOMY-20260923-v2.1.md) · -[Current workflow](../governance/OP-WORKFLOW-20260930.md) · +[Current authority](../GOVERNANCE.md) · +[Prior workflow](../governance/OP-WORKFLOW-20260930.md) · +[Historical delegation](../governance/OP-AUTONOMY-20260923-v2.1.md) · [Privacy rule](../governance/OP-PRIVACY-20260927.md) · [Home](../README.md) · [Historical material](../history/2025/README.md) From 63ae24f10de2e04c18c452923dff524d17b86cb7 Mon Sep 17 00:00:00 2001 From: Dylan Roy <249851642+d6g8k5htny-coder@users.noreply.github.com> Date: Mon, 9 Nov 2026 13:50:40 -0600 Subject: [PATCH 2/3] Repair historical Atlas profile control and run prepared Node regressions --- .github/workflows/cbz-node-regression.yml | 106 ++++++++++++++++++ DECISIONS.md | 2 + docs/cbz/NODE_FORMAL_FIXTURE_20261109.md | 86 ++++++++++++++ .../plans/2026-11-09-node-formal-fixture.md | 71 ++++++++++++ tests/test_atlas_formal_evidence.mjs | 23 +++- 5 files changed, 285 insertions(+), 3 deletions(-) create mode 100644 .github/workflows/cbz-node-regression.yml create mode 100644 docs/cbz/NODE_FORMAL_FIXTURE_20261109.md create mode 100644 docs/superpowers/plans/2026-11-09-node-formal-fixture.md diff --git a/.github/workflows/cbz-node-regression.yml b/.github/workflows/cbz-node-regression.yml new file mode 100644 index 000000000..7191389d5 --- /dev/null +++ b/.github/workflows/cbz-node-regression.yml @@ -0,0 +1,106 @@ +name: CBZ prepared Node regression + +'on': + pull_request: + push: + branches: [main] + workflow_dispatch: + +permissions: + contents: read + +jobs: + prepared-node: + runs-on: ubuntu-24.04 + timeout-minutes: 20 + defaults: + run: + shell: bash + working-directory: main + env: + PYTHONDONTWRITEBYTECODE: '1' + GIT_TERMINAL_PROMPT: '0' + steps: + - name: Check out complete Main history + uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 + with: + path: main + fetch-depth: 0 + persist-credentials: false + - name: Check out pinned public Math source + uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 + with: + repository: d6g8k5htny-coder/Math- + ref: f532c3e2d91a0c7679a37c733741660917a9b21b + path: Math- + fetch-depth: 0 + persist-credentials: false + - name: Check out pinned public query source + uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 + with: + repository: d6g8k5htny-coder/query- + ref: fb4c457b5861dc49696fd0a8e11c7c405056c05f + path: query- + fetch-depth: 0 + persist-credentials: false + - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 + with: + python-version: '3.11' + - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 + with: + node-version: '24' + - name: Prepare fresh evidence and immutable historical inputs + run: | + set -euo pipefail + cbz_evidence="$RUNNER_TEMP/cbz-node-regression-$GITHUB_RUN_ID-$GITHUB_RUN_ATTEMPT" + mkdir "$cbz_evidence" + python --version > "$cbz_evidence/python.txt" + node --version > "$cbz_evidence/node.txt" + git --version > "$cbz_evidence/git.txt" + git rev-parse HEAD > "$cbz_evidence/main-commit.txt" + git -C "$GITHUB_WORKSPACE/Math-" rev-parse HEAD > "$cbz_evidence/math-commit.txt" + git -C "$GITHUB_WORKSPACE/query-" rev-parse HEAD > "$cbz_evidence/query-commit.txt" + fetch_historical() { + git -c core.hooksPath=/dev/null -c credential.helper= -c core.fsmonitor=false \ + -c protocol.ext.allow=never -c maintenance.auto=false -c gc.auto=0 \ + -C "$GITHUB_WORKSPACE/$1" fetch --no-tags --no-write-fetch-head \ + --no-recurse-submodules --no-auto-maintenance origin "$2" + git -C "$GITHUB_WORKSPACE/$1" cat-file -e "$2^{commit}" + } + fetch_historical main 71400b94f6cb354a8cf7aba73ffede2138a64efa + fetch_historical main a12c178c0f857a130cf434e9efd44233a038195b + fetch_historical main f2e432ea5c86742e480c66624775bc9103343314 + fetch_historical Math- d6628da09384728992dcbe6e921cc28ba85aebb0 + fetch_historical Math- 82247833b5dd58e04291d55353b52938d73ec614 + fetch_historical Math- 9d7b6802424fb4715b31999066aafca8ee2f3cca + fetch_historical query- c88768bb11efd1f7d6bda188f13064bedec54a06 + - name: Admitted prepared-helper controls in both Python modes + run: | + set -euo pipefail + cbz_evidence="$RUNNER_TEMP/cbz-node-regression-$GITHUB_RUN_ID-$GITHUB_RUN_ATTEMPT" + python -B -S -m unittest -v tests.test_cbz_node_regression.NodeRunnerTests > "$cbz_evidence/helper-controls-normal.log" 2>&1 + python -B -O -S -m unittest -v tests.test_cbz_node_regression.NodeRunnerTests > "$cbz_evidence/helper-controls-optimized.log" 2>&1 + - name: Every tracked Node test with prepared fixtures + timeout-minutes: 5 + run: | + set -euo pipefail + python -B -S tools/cbz_node_regression.py \ + --math-root "$GITHUB_WORKSPACE/Math-" --query-root "$GITHUB_WORKSPACE/query-" \ + --output "$RUNNER_TEMP/cbz-node-regression-$GITHUB_RUN_ID-$GITHUB_RUN_ATTEMPT/normal" + - name: Every tracked Node test with optimized Python preparation + if: ${{ !cancelled() }} + timeout-minutes: 5 + run: | + set -euo pipefail + python -B -O -S tools/cbz_node_regression.py \ + --math-root "$GITHUB_WORKSPACE/Math-" --query-root "$GITHUB_WORKSPACE/query-" \ + --output "$RUNNER_TEMP/cbz-node-regression-$GITHUB_RUN_ID-$GITHUB_RUN_ATTEMPT/optimized" + - name: Preserve original results, TAP and complete fixture trees + if: always() + uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 + with: + name: cbz-node-regression-${{ github.run_id }}-${{ github.run_attempt }} + path: ${{ runner.temp }}/cbz-node-regression-${{ github.run_id }}-${{ github.run_attempt }} + include-hidden-files: true + if-no-files-found: error + retention-days: 30 diff --git a/DECISIONS.md b/DECISIONS.md index bc657fd59..d2bd54ce1 100644 --- a/DECISIONS.md +++ b/DECISIONS.md @@ -115,3 +115,5 @@ Append-only. Historical source outcomes remain in their original records. | 2026-11-09 18:30 | 2 | B/CI | The post-merge Workspace pins job passed eight network and strict-current pins, four canonical inputs and 187 tests. The post-merge federation workflow failed before any job with GitHub internal-server correlation `63dc5ffa-474e-4405-925b-574e391c43c6`; GitHub refused its failed-job retry with HTTP 403. Preserve both outcomes, do not infer a passing post-merge federation run, and continue other authorized checks. | Yes | Workspace pins run `42674111147`, job `129351557433`; federation run `42674110867`. | | 2026-11-09 17:32 | 2 | B/tooling | Selectively recovered the unchanged tracked-source inventory tool and retained tests from closed-unmerged Main #695, preserving original source authorship instead of creating a new final controller. Missing CLI exit 2 and retained missing-implementation behavioral failure exit 1 were observed first. CPython 3.12.14 then passed 75/75 controls in normal and optimized modes with no top-level skips, errors or failures; each suite exercised nine intended source mutants under the preserved classifier limits. Generate and verify exited 0 over nine explicit local federation checkouts, 27,826 tracked entries and eight gitlinks, with zero classifications. This checks source/state and declared-binding integrity only; recorded clean dirty flags do not award cleanliness, supplied receipt bytes do not authenticate kernel execution, and supplied quarantine strings do not prove isolation or satisfy the complete current governance contract. No historical catalogue, classification or isolation assertion is restored. CBZ flags remain false; no proof, bound, axiom, historical failure, kernel outcome or scientific status changed and no gate or criterion is awarded. Candidate full regression, committed-tree G11 and engineering review remain separate integration checks. | Yes; recoverable ordinary Git addition | Source head `82d9e480a51b946b3e74525ccf32139129d37d28`; recovery parent `5ebe5771d44160f56d4f2fd1e11726858d17dc90`; tool/test blobs `58f5cafbe151f7304ee9ea35b6fc46859d35630f` / `ac8e5edc832ae9ded7ecad134264d3bd1f1871c5`; inventory SHA-256 `c03fcaa8613ff3c853c1cfe0c50f82e05e7a0bb20207a3fe462a722eaab493d3`; normal/optimized focused log SHA-256 `7b3c3f3e6935b83ce071ed425c14bbe65d3d77157b039d3c2253fe0fbf3a3b54` / `35f7e951b52c0bb66bfba579c272c279c654cc506838efa2d365bc34659ebf57`; [recovery scope and commands](docs/cbz/INVENTORY_RECOVERY.md). | | 2026-11-09 17:51 | 2 | B/tooling | Repaired the inventory's real filter-configuration/status race after engineering review demonstrated harmless fixture command execution before a later refusal. A same-size real-Git regression failed against restored source with the executed marker, then passed after status was moved into private fixed detached Git metadata with a bounded regular index copy and explicit common objects. Preflight/postflight refusal remains, but source configuration no longer supplies status filter commands. Added linked-worktree rename/untracked, split-index refusal, ambient/source-info attribute and ignore policy, safe temporary-directory fallback and isolation-bypass mutation controls. The initial parent-directory guard was repaired RED→GREEN with an external `/var/tmp` fallback. CPython 3.12.14/Git 2.52.0 passed 81/81 focused controls per mode with no top-level skips/errors/failures; ten declared mutants were exercised per mode and source digests remained stable. Fresh repaired-tool generate/verify each exited 0 over nine unchanged parent checkouts and 27,826 entries, with zero classifications and false certification flags. Original recovery, reproduced failure, superseded development logs and stopped pre-repair candidate-wide runs remain preserved; no pass credit is assigned to stopped or superseded runs. No proof/catalogue/quarantine, bound, axiom, kernel outcome or scientific status changed; full candidate regression, committed-tree G11 and final federation gates remain separate. | Yes; separate ordinary Git repair from unchanged restoration `1e58f4d9fef8ea9016fc8e46e67a5f7dd33883dd` | Repaired tool/test blobs `2d100b241bc9d8bc6eb5ee6e9b16df229f41c447` / `9eb45d45961169e86c3469ba81f53f5f216bae84`; focused normal/optimized log SHA-256 `81f0dadc2830aa0818724e9c10c5056a2027338fc631fb9a6a045719eda4ffca` / `9058643300f8eb19cd559dfe0418283c7d1a203db6601e36bb781be9bd24eb47`; fresh parent inventory SHA-256 `c03fcaa8613ff3c853c1cfe0c50f82e05e7a0bb20207a3fe462a722eaab493d3`; [repair scope and evidence](docs/cbz/INVENTORY_RECOVERY.md#current-status-filter-isolation-repair--9-november-2026). | +| 2026-11-09 19:17 | 2 | B/testing | Replace one contradictory Node negative fixture premise with an explicit historical 8d43 profile control: require its exact source bridge, original 3b39 execution and REJECTED alignment, then change only manifest identity and require null. The three current.json versions checked in the ancestry of Main f1ef use the admitted 8d43 profile; this bounded search supplies no honest older-unmapped fixture. Preserve the original RED and current snapshot externally. Production model, bridge data, mathematics and historical records are unchanged. Focused/full prepared Node qualification and one intended-assertion guard mutant remain source-bound external observations; no final gate or scientific admission is inferred. | Yes | [Rationale and provenance](docs/cbz/NODE_FORMAL_FIXTURE_20261109.md); raw RED at Main `24ad8a143862fb9255fc6e2e7bf08c12f989f71d`: 767 tests, 766 passed, one failed, zero skipped/cancelled/todo; original TAP SHA-256 `f4fc26a2422b7bd85d4d5de951379b9b9a3365db2dbc147a7f8d88c7d5fc1f80`. | +| 2026-11-09 19:43 | 2 | B/testing, CI | Add an ordinary read-only prepared Node regression job because existing CI did not run the complete suite on this repaired test, and shared local PID/thread exhaustion prevents a new container qualification. Use exact public sibling pins, explicit historical fetches, Python 3.11/Node 24, admitted helper controls and sequential normal/optimized prepared runs with original artifacts retained on failure. Preserve prior RED and reviewed local evidence; require an actual hosted pass before merge, with no helper/model/math changes or privilege workaround. | Yes | [Design and provenance](docs/cbz/NODE_FORMAL_FIXTURE_20261109.md); reviewed local repair `94c8b2264eff0d32727126b7b027b2077ab81552`, composed source `aa4a74bff613b58224ea7995dbb0fa48c62d38b0`, actual Main parent `b1d36c80403acc96d257b9e71140a6cb83a595af`; `.github/workflows/cbz-node-regression.yml`. | diff --git a/docs/cbz/NODE_FORMAL_FIXTURE_20261109.md b/docs/cbz/NODE_FORMAL_FIXTURE_20261109.md new file mode 100644 index 000000000..6e6a9d3bf --- /dev/null +++ b/docs/cbz/NODE_FORMAL_FIXTURE_20261109.md @@ -0,0 +1,86 @@ +# Historical Atlas profile refusal control + +The prepared Node regression at Main `24ad8a143862fb9255fc6e2e7bf08c12f989f71d` +ran 767 tests: 766 passed and one failed, with no skipped, cancelled or todo +tests. The failure was the assertion that the checked-in `current.json` must +remain unmapped. Its raw TAP, result, exact current snapshot and source files +are retained outside the source tree under `node-regression-24ad` and +`node-formal-fixture-fix/baseline-red` in the session evidence. + +That assertion was introduced by `6edc839b57f0f06fc0748fd13cde820c7d49da42`. +Commit `2429c5fcd2eed458824a5cba71d09d7ef7be0e0f` later added an explicit +source bridge for the public-site `8d43` manifest and a positive test of the +same snapshot, claim and model. It left the older null assertion unchanged. +The two tests consequently required opposite results for identical input. + +The three committed versions of `docs/site/current.json` found by following +that path through the ancestry of Main +`f1ef3a2babecca0f0d4eb35b018dc81d30e9d62c` use +manifest blob `8d43188691c695f66e233727322484da22662a58`, 17,411 bytes. +The latest snapshot has blob `a64fb6d9768ada12b9884adac854e0ccdd5a910e`, +962,639 bytes and SHA-256 +`560d3d4e77369835c832dbb4fd0d12f4988744045b9cbc4c7978beddb6e9d20f`. +No genuine older unmapped snapshot was found in this bounded history. The +historical source is preserved; no invented older observation replaces it. + +The successor control starts from the existing pinned +`tests/fixtures/atlas_formal_projection.json` (SHA-256 +`7b36e2376f4010444108e81f316293ef3b3834d767e0900f68d684c760af2db1`). +It explicitly selects this historical projection profile: + +- Commit: `b7deb198722582df191006e4f47238f92e6e543d`. +- Manifest blob: `8d43188691c695f66e233727322484da22662a58`. +- Length: 17,411 bytes. +- Manifest SHA-256: `1c86e1eed110df3ef283f3a6e402194f82f682ba06e55e3ad8fe2010f152df7e`. + +This is a controlled profile fixture, not a claim that it is an unmapped +historical observation. The positive baseline must select the exact admitted +`8d43` bridge while retaining execution at +`3b39c42d6edcc98ae165270be346f0e31c57f81d` and the original `REJECTED` +alignment. The control then changes only the manifest blob identity and must +receive `null`. This retains a live refusal assertion rather than merely +inverting the obsolete expectation. + +Production model, bridge data, mathematical sources, original execution and +alignment records are unchanged. A targeted guard-removal mutant must fail +the new refusal assertion rather than crash. Focused controls and the full +prepared Node regression in normal and optimized host Python must be recorded +against the committed candidate. These finite operational observations grant +no new scientific status, Python 3.11 cold-start result or final CBZ gate. + +## Hosted prepared regression + +The reviewed repair was composed after the actual Main #780 merge +`b1d36c80403acc96d257b9e71140a6cb83a595af` as local commit +`aa4a74bff613b58224ea7995dbb0fa48c62d38b0`. Its original evidence and native +recovery bundle remain external and unchanged. Local process/thread exhaustion +also prevented an ordinary rootless container shim; no PID 1 modification or +privilege workaround is used to obtain a new execution result. + +The root-authorized follow-on adds `.github/workflows/cbz-node-regression.yml`. +Existing public-shop CI runs selected Node files, and the older cold-start +workflow's path filter omits this repaired Node test. The new workflow calls +the existing prepared helper over every tracked Node entry on ordinary pull +requests, Main pushes and manual dispatch, with read-only contents permission. +It uses an ordinary Ubuntu 24.04 runner, Python 3.11 and Node 24. Main has full +history under `main`; the exact public siblings are Math- +`f532c3e2d91a0c7679a37c733741660917a9b21b` and query- +`fb4c457b5861dc49696fd0a8e11c7c405056c05f`. The seven immutable historical +commits listed in `COLD_START.md` are fetched explicitly without moving HEAD. + +The admitted `NodeRunnerTests` class runs in both Python modes; the unadmitted +mutation/control launcher is not invoked. The actual prepared regression then +runs sequentially in normal and optimized Python with fresh mode directories +under `RUNNER_TEMP`, outside every source root. The optimized attempt runs +after a normal failure unless the job is cancelled; a failed command still +fails the job. An always-run artifact step retains the original result files, +TAP and complete fixture trees, including failures. Source/helper/model bytes +are not edited by the workflow. + +This small dedicated job was chosen over expanding the unrelated navigation +suite or invoking the unadmitted cold-start control harness. Local validation +is structural YAML parsing with the already installed parser; there are no +new mirror tests or package installations. Root publishes the candidate, and +its actual hosted job must pass before merge. No future hosted outcome is +inferred from the prior 767/767 local results or the intended guard-mutant +failure. diff --git a/docs/superpowers/plans/2026-11-09-node-formal-fixture.md b/docs/superpowers/plans/2026-11-09-node-formal-fixture.md new file mode 100644 index 000000000..ae08303af --- /dev/null +++ b/docs/superpowers/plans/2026-11-09-node-formal-fixture.md @@ -0,0 +1,71 @@ +# Historical Atlas profile control implementation plan + +> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking. + +> Execute the root-approved bounded test repair in the separate +> `cbz/node-formal-fixture-20261109` branch. No remote writes are authorized. + +**Goal:** replace one contradictory negative fixture premise with an explicit +historical-profile baseline and a manifest-identity refusal assertion. + +**Architecture:** reuse the pinned projection fixture and the production +`formalConnection` unchanged. Select the exact admitted `8d43` profile, check +historical execution and rejected alignment, then change only its manifest +blob identity and require refusal. + +**Tech stack:** existing Node test runner and prepared Python regression helper. + +**Spec:** `docs/cbz/NODE_FORMAL_FIXTURE_20261109.md` and the root's explicit +bounded repair authorization following the preserved `24ad` RED observation. + +## Constraints and review focus + +- Base is Main `f1ef3a2babecca0f0d4eb35b018dc81d30e9d62c`. +- Keep `cbz-qualified-next`, its full regressions and `.lake` untouched. +- Preserve the original RED and actual current snapshot outside source. +- Change no production model, source bridge, mathematics or tool registry. +- Retain explicit refusal coverage and the original historical execution and + adverse alignment; do not fabricate an older unmapped record. + +## Task: repair and qualify the fixture contract + +This is the approved sequence; actual observations are recorded outside source +against the committed candidate SHA, rather than treating this plan as a status +register. + +- [ ] Preserve and bind the observed 767/766/1 RED and native source identities. +- [ ] Replace only the obsolete test with the explicit profile control. +- [ ] Record rationale/provenance here and append the existing decision ledger. +- [ ] Run the focused evidence/bridge Node files with zero failures/skips. +- [ ] Commit the candidate locally before prepared endpoint qualification. +- [ ] Run the actual full prepared Node suite in normal and optimized host + Python; require nonempty TAP, tests=pass and fail/cancelled/skipped/todo=0. +- [ ] In an external disposable source copy, remove only the unknown-manifest + refusal guard. The selected new test must fail its intended null assertion, + with a successful unmutated control and no crash/import error. +- [ ] Review the complete diff and evidence, preserving all source endpoints and + failed observations. Root reviews and publishes separately. + +## Follow-on: expose the prepared suite in ordinary CI + +Root authorized this small CI follow-on after the reviewed local composition +`aa4a74bff613b58224ea7995dbb0fa48c62d38b0`, whose sole parent is actual Main +merge `b1d36c80403acc96d257b9e71140a6cb83a595af`. Preserve those commits, +the original `94c8` qualification and all failed setup observations. + +**Files:** add `.github/workflows/cbz-node-regression.yml`; update this plan, +the existing rationale and append one decision row. The repaired test and all +helper/model/math files retain their reviewed bytes. + +- [ ] Add one ordinary pull-request/Main-push/manual job with read-only contents + permission, full-history Main under `main`, pinned public Math/query + siblings, Python 3.11 and Node 24. Use existing immutable action pins. +- [ ] Fetch the seven explicit historical commits from `COLD_START.md`, then + run only the admitted helper self-test class in normal/optimized modes. +- [ ] Call the existing prepared helper sequentially in both modes with fresh + external `RUNNER_TEMP` directories and fail-closed shell exits; preserve + the original results, TAP and complete fixture trees with `always()`. +- [ ] Parse the YAML with the installed parser and inspect its structure; add + no mirror tests, packages, privilege workarounds or unadmitted harness. +- [ ] Commit locally with exact tree/parent and source-blob identities. Root + publishes via API; require the actual hosted job to pass before merge. diff --git a/tests/test_atlas_formal_evidence.mjs b/tests/test_atlas_formal_evidence.mjs index ebd62a9f7..700649fe8 100644 --- a/tests/test_atlas_formal_evidence.mjs +++ b/tests/test_atlas_formal_evidence.mjs @@ -38,9 +38,26 @@ test('An equal-looking different node ID is never inferred to be covered',()=>{ test('Other actual claims keep missing mappings explicit',()=>{ assert.equal(formalConnection(fixture(),'math.uniform-matrix-cap-lifetime'),null); }); -test('The actual older committed observation has a different manifest and remains unmapped',()=>{ - const old=JSON.parse(fs.readFileSync(new URL('../docs/site/current.json',import.meta.url))); - assert.equal(formalConnection(old,id),null); +test('An explicit historical 8d43 profile rejects a changed manifest identity',()=>{ + // Controlled profile derived from the pinned fixture, not an older observed snapshot. + const data=fixture(); + const manifest=data.sources.find(s=>s.repository===EVIDENCE.formal.repository + &&s.path===EVIDENCE.formal.manifest_path); + Object.assign(manifest,{commit:'b7deb198722582df191006e4f47238f92e6e543d', + blob:'8d43188691c695f66e233727322484da22662a58',bytes:17411}); + const connection=formalConnection(data,id); + assert.ok(connection,'The explicit 8d43 profile must use its admitted historical source bridge'); + assert.equal(connection.source_bridge.projection_manifest.commit,'b7deb198722582df191006e4f47238f92e6e543d'); + assert.equal(connection.source_bridge.projection_manifest.blob,'8d43188691c695f66e233727322484da22662a58'); + assert.equal(connection.source_bridge.projection_manifest.bytes,17411); + assert.equal(connection.source_bridge.record_sha256,'8cd6d4bcb48874aa40a3aa1d7e36d113487a43a1608e24fd4023cbb9e0f64299'); + assert.equal(connection.formal.commit,'3b39c42d6edcc98ae165270be346f0e31c57f81d'); + assert.equal(connection.execution.checked_commit,'3b39c42d6edcc98ae165270be346f0e31c57f81d'); + assert.equal(connection.alignment.record.disposition,'REJECTED'); + assert.equal(connection.alignment.record.manifest_sha256,'8dfcfe4e9296bca47eede517f37b546ea420ec6b51944b7c16164a906b40e090'); + manifest.blob='0'.repeat(40); + assert.equal(formalConnection(data,id),null, + 'An unrecognized manifest identity must not attach historical evidence'); }); test('An execution for another revision cannot attach to the artifact',()=>{ const e=structuredClone(EVIDENCE);e.execution.checked_commit='0'.repeat(40); From 962923a12f82e88f54ece180fbabdf05934aae78 Mon Sep 17 00:00:00 2001 From: Dylan Roy <249851642+d6g8k5htny-coder@users.noreply.github.com> Date: Mon, 9 Nov 2026 14:00:45 -0600 Subject: [PATCH 3/3] Expose source-bound Node regression outcomes in CI logs --- .github/workflows/cbz-node-regression.yml | 19 +++++++++++++++++++ DECISIONS.md | 1 + docs/cbz/NODE_FORMAL_FIXTURE_20261109.md | 6 ++++++ 3 files changed, 26 insertions(+) diff --git a/.github/workflows/cbz-node-regression.yml b/.github/workflows/cbz-node-regression.yml index 7191389d5..2dbc420b1 100644 --- a/.github/workflows/cbz-node-regression.yml +++ b/.github/workflows/cbz-node-regression.yml @@ -95,6 +95,25 @@ jobs: python -B -O -S tools/cbz_node_regression.py \ --math-root "$GITHUB_WORKSPACE/Math-" --query-root "$GITHUB_WORKSPACE/query-" \ --output "$RUNNER_TEMP/cbz-node-regression-$GITHUB_RUN_ID-$GITHUB_RUN_ATTEMPT/optimized" + - name: Report bounded observed outcomes + if: ${{ !cancelled() }} + run: | + python -B -S - "$RUNNER_TEMP/cbz-node-regression-$GITHUB_RUN_ID-$GITHUB_RUN_ATTEMPT" <<'PY' + import json, pathlib, sys + root = pathlib.Path(sys.argv[1]) + for mode in ('normal', 'optimized'): + controls = root / f'helper-controls-{mode}.log' + if controls.is_file(): + print('HELPER CONTROLS ' + mode) + print('\n'.join(controls.read_text(encoding='utf-8').splitlines()[-4:])) + report = root / mode / 'result.json' + if report.is_file(): + record = json.loads(report.read_text(encoding='utf-8')) + keys = ('schema', 'status', 'summary', 'node_exit', 'source_before', 'source_after', 'fixture_files', 'fixture_tree_sha256', 'log_sha256', 'node_version') + print(json.dumps({'mode': mode, **{key: record.get(key) for key in keys}}, sort_keys=True)) + else: + print(json.dumps({'mode': mode, 'report_missing': True})) + PY - name: Preserve original results, TAP and complete fixture trees if: always() uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 diff --git a/DECISIONS.md b/DECISIONS.md index d2bd54ce1..f4da61aa4 100644 --- a/DECISIONS.md +++ b/DECISIONS.md @@ -117,3 +117,4 @@ Append-only. Historical source outcomes remain in their original records. | 2026-11-09 17:51 | 2 | B/tooling | Repaired the inventory's real filter-configuration/status race after engineering review demonstrated harmless fixture command execution before a later refusal. A same-size real-Git regression failed against restored source with the executed marker, then passed after status was moved into private fixed detached Git metadata with a bounded regular index copy and explicit common objects. Preflight/postflight refusal remains, but source configuration no longer supplies status filter commands. Added linked-worktree rename/untracked, split-index refusal, ambient/source-info attribute and ignore policy, safe temporary-directory fallback and isolation-bypass mutation controls. The initial parent-directory guard was repaired RED→GREEN with an external `/var/tmp` fallback. CPython 3.12.14/Git 2.52.0 passed 81/81 focused controls per mode with no top-level skips/errors/failures; ten declared mutants were exercised per mode and source digests remained stable. Fresh repaired-tool generate/verify each exited 0 over nine unchanged parent checkouts and 27,826 entries, with zero classifications and false certification flags. Original recovery, reproduced failure, superseded development logs and stopped pre-repair candidate-wide runs remain preserved; no pass credit is assigned to stopped or superseded runs. No proof/catalogue/quarantine, bound, axiom, kernel outcome or scientific status changed; full candidate regression, committed-tree G11 and final federation gates remain separate. | Yes; separate ordinary Git repair from unchanged restoration `1e58f4d9fef8ea9016fc8e46e67a5f7dd33883dd` | Repaired tool/test blobs `2d100b241bc9d8bc6eb5ee6e9b16df229f41c447` / `9eb45d45961169e86c3469ba81f53f5f216bae84`; focused normal/optimized log SHA-256 `81f0dadc2830aa0818724e9c10c5056a2027338fc631fb9a6a045719eda4ffca` / `9058643300f8eb19cd559dfe0418283c7d1a203db6601e36bb781be9bd24eb47`; fresh parent inventory SHA-256 `c03fcaa8613ff3c853c1cfe0c50f82e05e7a0bb20207a3fe462a722eaab493d3`; [repair scope and evidence](docs/cbz/INVENTORY_RECOVERY.md#current-status-filter-isolation-repair--9-november-2026). | | 2026-11-09 19:17 | 2 | B/testing | Replace one contradictory Node negative fixture premise with an explicit historical 8d43 profile control: require its exact source bridge, original 3b39 execution and REJECTED alignment, then change only manifest identity and require null. The three current.json versions checked in the ancestry of Main f1ef use the admitted 8d43 profile; this bounded search supplies no honest older-unmapped fixture. Preserve the original RED and current snapshot externally. Production model, bridge data, mathematics and historical records are unchanged. Focused/full prepared Node qualification and one intended-assertion guard mutant remain source-bound external observations; no final gate or scientific admission is inferred. | Yes | [Rationale and provenance](docs/cbz/NODE_FORMAL_FIXTURE_20261109.md); raw RED at Main `24ad8a143862fb9255fc6e2e7bf08c12f989f71d`: 767 tests, 766 passed, one failed, zero skipped/cancelled/todo; original TAP SHA-256 `f4fc26a2422b7bd85d4d5de951379b9b9a3365db2dbc147a7f8d88c7d5fc1f80`. | | 2026-11-09 19:43 | 2 | B/testing, CI | Add an ordinary read-only prepared Node regression job because existing CI did not run the complete suite on this repaired test, and shared local PID/thread exhaustion prevents a new container qualification. Use exact public sibling pins, explicit historical fetches, Python 3.11/Node 24, admitted helper controls and sequential normal/optimized prepared runs with original artifacts retained on failure. Preserve prior RED and reviewed local evidence; require an actual hosted pass before merge, with no helper/model/math changes or privilege workaround. | Yes | [Design and provenance](docs/cbz/NODE_FORMAL_FIXTURE_20261109.md); reviewed local repair `94c8b2264eff0d32727126b7b027b2077ab81552`, composed source `aa4a74bff613b58224ea7995dbb0fa48c62d38b0`, actual Main parent `b1d36c80403acc96d257b9e71140a6cb83a595af`; `.github/workflows/cbz-node-regression.yml`. | +| 2026-11-09 20:00 | 2 | B/testing, CI | Print bounded fields from original prepared-Node results and the helper-control log footers in CI, so counts and source/fixture identities are directly inspectable even when the local artifact reader cannot start. Preserve all original artifacts and grading behavior; no result is synthesized from a PASS line. | Yes | First hosted run `42687230494`, job `129397179313`, at head `63ae24f10de2e04c18c452923dff524d17b86cb7` succeeded and uploaded artifact `13705427515`; exact artifact case counts were not yet independently read. [Prepared regression contract](docs/cbz/NODE_FORMAL_FIXTURE_20261109.md). | diff --git a/docs/cbz/NODE_FORMAL_FIXTURE_20261109.md b/docs/cbz/NODE_FORMAL_FIXTURE_20261109.md index 6e6a9d3bf..96bdf951f 100644 --- a/docs/cbz/NODE_FORMAL_FIXTURE_20261109.md +++ b/docs/cbz/NODE_FORMAL_FIXTURE_20261109.md @@ -84,3 +84,9 @@ new mirror tests or package installations. Root publishes the candidate, and its actual hosted job must pass before merge. No future hosted outcome is inferred from the prior 767/767 local results or the intended guard-mutant failure. + +The job also prints bounded fields from each original result: actual TAP +summary, Node exit, before/after source commit, fixture count/tree identity and +TAP digest. It prints only the helper-control log footer, preserving the full +original logs in the artifact. This exposes observed outcomes without changing +the helper's grading or inferring a case count from a PASS line.