Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 10 additions & 0 deletions BASELINE.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,15 @@
# CBZ baseline ledger

## Main #807 and G11 recomposition boundary

Main #807 integrates scoped SDGA fixture replay tooling with qualification and classification BLOCKED and no scientific acceptance. The earlier expanded G11 census passed on its own tree; the combined tree requires fresh evidence and the final federation gates.


## G11 Python subprocess option limitation

Three actual synthetic subprocess controls exposed nonexecuting commands that the prior static census nevertheless credited. The source-bound repair and adverse tests are recorded in [STATUS](STATUS.md#g11-python-subprocess-operand-boundary--10-november-2026); the earlier census outcome retains this limitation. Local component checks do not establish final federation G11 or Layer A clearance.


## Interpreter attribution evidence boundary

The earlier exact-candidate G11 PASS is preserved with its later demonstrated interpreter-option attribution limitation. Eighteen bounded workflow controls and nineteen format mutations passed per mode on the recorded debug source; the original 328-case run then stopped at a missing target. One mutation target was adapted and its earlier definition retained. Current composed and final federation G11 evidence remains required.
Expand Down
206 changes: 206 additions & 0 deletions DECISIONS.md

Large diffs are not rendered by default.

7 changes: 6 additions & 1 deletion GOVERNANCE.md
Original file line number Diff line number Diff line change
@@ -1,13 +1,18 @@
# CBZ-0 governance

## G11 subprocess operand scope

Python test subprocess references use an explicit operand whitelist: Python `--check-hash-based-pycs` accepts `default`, `always` or `never`; `-W` accepts the six documented simple actions; `-X` accepts `dev`. Other option operands, including shell operands, are not credited. This is a bounded static analysis and does not establish execution. The actual adverse failures, earlier mutation survivors, adapted mutation definitions and local source-bound outcomes are retained in STATUS and native evidence; hosted and final federation gates remain separate.


## G11 interpreter-option boundary

The workflow census only credits bounded interpreter command forms. It refuses Python and shell options with unvalidated operands and does not treat version, help, alternate-code or syntax-only options as script execution. The exact commands and actual source-bound test, mutation and failure records are in STATUS and TOOLING. Static references do not establish execution. The earlier 328 mutation names remain, with one shell `-c` target adapted and its previous bytes preserved. The full current-source and federation gates remain separate.


## Current G11 workflow and origin boundary

The bounded G11 origin verifier records actual pre-cleanup frames and lifecycle state; `tools/cbz/g11_origin_mutation_selftest.py` tests seven origin-guard mutations. `tools/cbz/g11_workflow_formats_mutation_selftest.py` tests nineteen bounded env/JSON and interpreter-option workflow-scanner mutations. Both commands and their normal/optimized modes are in STATUS and TOOLING. The strict JSON path recognizes job-step runs and basic shapes, rejects invalid or duplicate-key input, and reports JSON no-op/parse findings with line 0 when position is unavailable. It is not a full GitHub workflow validator or a proof of execution. The latest per-source executions and failures are retained in STATUS and native evidence; final federation qualification remains open.
The bounded G11 origin verifier records actual pre-cleanup frames and lifecycle state; `tools/cbz/g11_origin_mutation_selftest.py` tests seven origin-guard mutations. `tools/cbz/g11_workflow_formats_mutation_selftest.py` tests 22 bounded workflow and Python-subprocess reference mutations; two earlier definitions were adapted and preserved. Both commands and their normal/optimized modes are in STATUS and TOOLING. The strict JSON path recognizes job-step runs and basic shapes, rejects invalid or duplicate-key input, and reports JSON no-op/parse findings with line 0 when position is unavailable. It is not a full GitHub workflow validator or a proof of execution. The latest per-source executions and failures are retained in STATUS and native evidence; final federation qualification remains open.

## G11 mutation-origin guard: current scope

Expand Down
10 changes: 10 additions & 0 deletions PROOF_INDEX.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,15 @@
# CBZ proof and quarantine index

## Main #807 tooling boundary

The SDGA fixture replay integration and G11 recomposition do not close, refute or rebind mathematical proof statements or establish scientific acceptance. Qualification and classification remain BLOCKED.


## G11 subprocess attribution boundary

The newly observed static-reference defect and its tooling repair do not close, refute or rebind mathematical statements or kernel outcomes, and they do not establish independent scientific acceptance.


## Interpreter-option tooling boundary

The G11 interpreter-option findings, repairs and retained failures concern tool-reference attribution only; they do not close, refute or rebind mathematical statements or establish scientific review.
Expand Down
29 changes: 29 additions & 0 deletions STATUS.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,34 @@
# CBZ-0 execution status

## Main #807 composition and current observations — 10 November 2026

Main #807 merged as `cff23146a56949e03cad4b15bdc4f07feba665ae`, tree `a39aa5a8064aa884446d3fe3884113c6111a60bd`, after the prior G11 candidate. It integrates guarded SDGA fixture replay tooling; its qualification and classification remain BLOCKED, execution authentication remains false, and no Task D, whole-kernel or scientific acceptance follows. The complete peer tree and its decision-ledger prefix are preserved in this composition.

The separately observed Workspace default `a4c56deaa4a0e69ec61dee3efc62da14a929a3bd` still pins Main `1a6acb192f0946900c4d445ee36febec27540da0`; after #807 it is stale and requires a fresh pin update. These 06:54 UTC reads were not atomic. The earlier expanded G11 candidate `d51700eb5466158206c3431f6b99c4db036fdc95` passed hosted census run `42771301709` in both modes: 1,508 scanned files, 6,078 tracked leaves and zero reported findings, failing tools or skips. Its original ZIP was independently checked and read back from Git blob `d2442c1a996aed2da6144e70a2bebfab13298e5f`. That census belongs to the earlier tree; this recomposition requires a fresh current-tree census and the final G0–G11 pass.

Main `1a6acb19` site run `42768238849` failed public-history readback, its browser evidence upload failed for the absent skipped output directory, and deployment was skipped. A later authenticated read-only run `42771407975` returned HTTP 200 with an empty prepared Pages deployment status; the separately read receipt ref remained `6e01d1f5f83d320172e29391eea336e3256aeea2`. Empty status does not establish terminal cancellation, recovery or publication. Original observations are retained under `cbz/evidence/pages-prepared-current-20261110/`.


## G11 subprocess evidence and expanded controls — 10 November 2026

The earlier candidate `ccb98a32b83b3dc82197ec52efd64ff44f622865` passed hosted Python 3.11 component push and PR runs `42770025870` and `42770042252`, each with 265 census controls, 12 origin controls, seven origin mutations, 22 workflow/subprocess controls, 22 workflow/subprocess mutations and 328 original mutations per mode. All 714 mutant records per run were checked for designated authorized origin; original mutation ZIPs and their byte-identical Git readbacks are retained in [evidence](cbz/evidence/g11-python-subprocess-options-20261110/). Its exact-tree hosted census `42770128530` passed both modes with 1,508 scanned files, 6,055 tracked leaves and zero reported findings, failing tools or skips. These are source-bound observations of the earlier tree, not final federation G11.

The control suite now exercises all 17 supported separated or attached Python option combinations using actual marker-bearing subprocesses, and adds distinct invalid separate `-X` and `sh -o` controls. Local CPython 3.12.14 ran 287 census/format controls and 22 designated mutations per mode, without failures or skips; the original 328-case and origin campaigns from the unchanged census, original test and harness remain the earlier local observations. Fresh hosted checks and an exact-tree census are required for this expanded source.

The first protected workflow `42770042202` failed during architecture preparation because the runner daemon configuration was marked unsafe/private; architecture controls were skipped and the required aggregate failed on that dependency. Original logs and artifact metadata are retained; the potentially private original artifact is not copied to public Git. A failed-job retry ran separately. No earlier formal success is silently rebound to a different attempt.


## G11 Python subprocess operand boundary — 10 November 2026

On Main `1a6acb192f0946900c4d445ee36febec27540da0`, local actual subprocess controls found that the earlier static census credited three Python-test commands despite nonzero exits and absent execution markers: invalid Python `--check-hash-based-pycs`, invalid attached `-Xutf8=2`, and invalid Bash `-o` operands. The [normal and optimized RED logs](cbz/evidence/g11-python-subprocess-options-20261110/) each contain three selected assertion failures and one passing supported-option control on CPython 3.12.14. The earlier exact-tree census outcome remains historical with this synthetic-fixture attribution limitation; it is not final G11 or a claim that any particular production tool was unexercised.

This source bounds Python subprocess interpreter operands to the documented Python options and values and declines shell options requiring unvalidated operands. It retains static-reference semantics; a reference does not prove actual execution. On the earlier local candidate `ccb98a32b83b3dc82197ec52efd64ff44f622865`, `python3 -B -S -m unittest tests.test_g11_census tests.test_g11_workflow_formats` and its `-O` variant each ran 285 tests without failures or skips. The workflow/subprocess mutation command ran 22 cases in each mode with authorized selected-body failures; the separate unchanged original campaign ran 328 cases in each mode, with 263 clean baseline tests and 328 authorized kills. Twelve origin controls per mode passed. These local Python 3.12 observations and [their evidence](cbz/evidence/g11-python-subprocess-options-20261110/) do not replace hosted Python 3.11 checks, a current-source census or the final federation pass.

The earlier 19 workflow mutation names remain. Two definitions were adapted because the new central guard made the old single-guard mutations survive; their previous source and both 17-of-19 outcomes are preserved. Three subprocess mutations were added. An initial partial local fixture lacked three native workflows and produced three errors; the reproduced failure is retained and the workflows were subsequently recovered from exact Git blobs. No failure is silently discarded.

Main #804 merged as `1a6acb192f0946900c4d445ee36febec27540da0` with its tested tree; Workspace #73 merged as `a4c56deaa4a0e69ec61dee3efc62da14a929a3bd` and pinned that Main commit. Its postmerge pin and contract jobs succeeded. These observations are separately timed, not final G7 or CBZ-0.


## Main advanced after the earlier candidate — 10 November 2026

Main #806 merged as `d291b42801a26252a046d8ca78a6bd62abc448ae`, advancing the separately observed default from `964390a94c08f00f119dc3b2e0be16180c1544e2`. Workspace `47221adf89749dc226e8e3578af2f8fd7c16ea20` still pins the latter at the separately observed cut; a fresh pointer update and strict-current checks are required. The earlier #804 exact-head census on `709975c6ff42f0bca020abd4164a45479e29a0df` passed in both modes with no reported findings, unexercised tools or skips; its [original verified ZIP and logs](cbz/evidence/g11-sealed-candidate-census-20261110/) are retained. The 06:04 non-atomic Workspace pin observation also predates #806. These receipts retain their original sources and do not establish qualification of this recomposition, final G7, scientific review or CBZ-0.
Expand Down
18 changes: 16 additions & 2 deletions TOOLING.md
Original file line number Diff line number Diff line change
@@ -1,8 +1,22 @@
# CBZ mission-tool registry

## Python subprocess operand validation

Purpose: prevent unsupported or invalid interpreter-option operands in Python subprocess references from manufacturing a static tool-reference credit. The implementation supports only Python `--check-hash-based-pycs` values `default`, `always`, `never`; `-W` values `default`, `error`, `ignore`, `always`, `module`, `once`; and `-X dev`, in separate or supported attached form. Shell options with operands and other values are conservatively uncredited. Supported references are not execution receipts.

```sh
python3 -B -S -m unittest tests.test_g11_census tests.test_g11_workflow_formats
python3 -B -S -O -m unittest tests.test_g11_census tests.test_g11_workflow_formats
python3 -B -S tools/cbz/g11_workflow_formats_mutation_selftest.py
python3 -B -S -O tools/cbz/g11_workflow_formats_mutation_selftest.py
```

Local CPython 3.12.14 executed 287 controls and 22 designated mutations per mode after expanding the positive and negative controls; the earlier source-bound 285-control results remain preserved. The original 328-case self-test also completed per mode; these are scoped local outcomes, not a final G11 certificate. Actual failures and two adapted earlier mutation definitions are preserved in [evidence](cbz/evidence/g11-python-subprocess-options-20261110/).


## G11 bounded interpreter handling

The workflow reference scanner reuses the bounded argument parser and refuses Python or shell interpreter options whose operands are not validated. Version/help, alternate-code and syntax-only controls are tested using actual subprocesses and no-execution markers. The 18-method workflow-format suite and 19-case mutation commands appear below. A static reference is not an execution receipt; unsupported option forms fail closed.
The workflow reference scanner reuses the bounded argument parser and refuses Python or shell interpreter options whose operands are not validated. Version/help, alternate-code and syntax-only controls are tested using actual subprocesses and no-execution markers. The earlier 18-method workflow-format suite and 19-case mutation commands are retained below; the current suite has 24 methods and 22 mutations. A static reference is not an execution receipt; unsupported option forms fail closed.


## G11 workflow-format mutation verification
Expand All @@ -16,7 +30,7 @@ env -u PYTHONPATH -u PYTHONHOME -u PYTHONPYCACHEPREFIX -u PYTHONSTARTUP -u PYTHO
env -u PYTHONPATH -u PYTHONHOME -u PYTHONPYCACHEPREFIX -u PYTHONSTARTUP -u PYTHONUSERBASE python3 -B -S -O tools/cbz/g11_workflow_formats_mutation_selftest.py
```

The nineteen declared mutations require clean selected baselines and authorized selected-body failures; the source-bound 19-case execution and its limits are recorded in STATUS. The actual old-census failure, red cases and bounded test results are retained under `cbz/evidence/g11-composed-census-20261110/` and `cbz/evidence/g11-workflow-formats-*`. Source references are static evidence, not proof of workflow execution; full JSON workflow semantics, arbitrary shell and dynamic conditions are not certified. JSON no-op and invalid-input findings use line 0 when their exact source line is unavailable.
The 22 current mutations require clean selected baselines and authorized selected-body failures; the earlier source-bound 19-case execution and its limits remain recorded in STATUS. The actual old-census failure, red cases and bounded test results are retained under `cbz/evidence/g11-composed-census-20261110/` and `cbz/evidence/g11-workflow-formats-*`. Source references are static evidence, not proof of workflow execution; full JSON workflow semantics, arbitrary shell and dynamic conditions are not certified. JSON no-op and invalid-input findings use line 0 when their exact source line is unavailable.

## G11 assertion-origin mutation verification

Expand Down
Loading
Loading