Downstream hard gate: fail-closed promotion control for main #90/#86

.github/workflows/downstream-gate.yml

@@ -0,0 +1,82 @@
+name: Downstream hard-gate exact replay
+'on':
+  push:
+    branches: [cursor/downstream-hard-gate-91fa]
+  pull_request:
+    paths: ['frontiers/downstream_gate_20260925/**', '.github/workflows/downstream-gate.yml']
+  workflow_dispatch:
+permissions:
+  contents: read
+jobs:
+  replay:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    steps:
+      - uses: actions/checkout@11d5960a326750d5838078e36cf38b85af677262
+      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
+        with:
+          python-version: '3.11'
+      - name: Verify sources and run gate controls plus regressions
+        env:
+          PYTHONDONTWRITEBYTECODE: '1'
+        run: |
+          python -B -S - <<'PY'
+          import hashlib, json, os, pathlib, subprocess, sys
+          root = pathlib.Path('frontiers/downstream_gate_20260925').resolve()
+          pins = json.loads((root / 'SOURCE_FILES.json').read_text())['files']
+          def verify():
+              for name, identity in pins.items():
+                  if pathlib.PurePosixPath(name).name != name or (root / name).is_symlink():
+                      raise ValueError('flat regular source required')
+                  raw = (root / name).read_bytes()
+                  if len(raw) != identity['bytes'] or hashlib.sha256(raw).hexdigest() != identity['sha256']:
+                      raise ValueError('source identity mismatch: ' + name)
+          if sys.version_info[:2] != (3, 11):
+              raise RuntimeError('actual Python 3.11 required')
+          verify()
+          out = pathlib.Path(os.environ['RUNNER_TEMP']) / 'downstream-gate-results'
+          subprocess.run([sys.executable, '-B', '-S', str(root / 'run_validation.py'), '--output', str(out)], check=True, timeout=120)
+          regressions = [
+              ('coefficients/side24_v1', 30),
+              ('frontiers/three_fronts_20260924', 57),
+              ('frontiers/price_budget_20260924', 26),
+              ('frontiers/full_price_20260924', 36),
+              ('frontiers/remote_window_20260924', 28),
+          ]
+          for mode, flags in [('normal', []), ('optimized', ['-O'])]:
+              for index, (folder, count) in enumerate(regressions):
+                  run = subprocess.run(
+                      [sys.executable, '-B', *flags, '-S', '-m', 'unittest', 'discover', '-s', folder, '-p', 'test_*.py', '-v'],
+                      capture_output=True, text=True, timeout=90,
+                  )
+                  (out / f'regression_{mode}_{index}.stdout').write_text(run.stdout)
+                  (out / f'regression_{mode}_{index}.stderr').write_text(run.stderr)
+                  if run.returncode or f'Ran {count} tests' not in run.stderr or 'skipped=' in run.stderr:
+                      raise RuntimeError('regression failed: ' + folder)
+          verify()
+          report = json.loads((out / 'REPORT.json').read_text())
+          if not report['passed'] or report['distinct_tests'] != 27 or report['distinct_semantic_mutations'] != 7:
+              raise RuntimeError('unexpected new-test scope')
+          (out / 'GITHUB_CONTEXT.json').write_text(json.dumps({
+              'run': os.environ['GITHUB_RUN_ID'],
+              'sha': os.environ['GITHUB_SHA'],
+              'python': sys.version,
+              'new_distinct_tests': 27,
+              'regression_distinct_tests': 177,
+              'total_distinct_tests': 204,
+              'new_semantic_mutations': 7,
+              'source_files_verified': len(pins),
+              'lemma_closed': False,
+              'scientific_effect': 'NONE',
+              'meaning': 'same-author integrity-gate checks; not analytic review',
+          }, indent=2) + '\n')
+          print('DOWNSTREAM_GATE_PASSED')
+          PY
+      - name: Retain evidence even on failure
+        if: always()
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
+        with:
+          name: downstream-gate-${{ github.run_id }}
+          path: ${{ runner.temp }}/downstream-gate-results
+          if-no-files-found: error
+          retention-days: 30

README.md

@@ -14,6 +14,7 @@
 | P15 failed extension | [Price-range counterexample](frontiers/three_fronts_20260924/P15_PRICE_BOUNDARY.md) | Unrestricted same-palette transformed-price extension is false |
 | P15 valid price extension | [Low-probability transformed-price budget](frontiers/price_budget_20260924/PROOF.md) | New sufficient hypotheses: demand at least 2 and probabilities at most 1/4 |
 | P15 full probability range | [Full transformed-price budget and sharp constant](frontiers/full_price_20260924/PROOF.md) | Demands at least 2; every independent probability vector; same palette; sharp uniform factor 1/[3-log(3e-2)] |
+| Downstream hard gate | [Fail-closed promotion control](frontiers/downstream_gate_20260925/README.md) | [main #90](https://github.com/d6g8k5htny-coder/main/issues/90) / [#86](https://github.com/d6g8k5htny-coder/main/issues/86); integrity only; `lemma_closed` stays false |
 
 All positive research claims here are author-side candidates pending nonauthor review. A merge publishes a source; a test checks its stated coverage. Neither independently accepts a theorem. The counterexample and the restricted successor concern different hypotheses. The full-range successor removes the probability cutoff, not the demand or realized-family hypotheses; the older 16/27 factor remains better on its smaller domain.
 
@@ -31,6 +32,8 @@ python -B -S frontiers/price_budget_20260924/price_budget.py
 python -B -S -m unittest discover -s frontiers/price_budget_20260924 -p 'test_*.py' -v
 python -B -S frontiers/full_price_20260924/full_price.py
 python -B -S -m unittest discover -s frontiers/full_price_20260924 -p 'test_*.py' -v
+python -B -S frontiers/downstream_gate_20260925/hard_gate.py
+python -B -S -m unittest discover -s frontiers/downstream_gate_20260925 -p 'test_*.py' -v
 ```
 
 The full-price package has 36 distinct tests and seven semantic mutation controls. Run `python -B -S frontiers/full_price_20260924/run_validation.py --output /tmp/full-price-new-run` for both modes; choose a new output directory outside the source tree. `--mode normal` or `--mode optimized` permits bounded split runs. Finite probability grids are checks, not the proof of the full-cube statement.

frontiers/README.md

@@ -7,5 +7,7 @@
 | Three fronts, 24 September 2026 | [Package overview](three_fronts_20260924/README.md) | Lifetime remainder, RN interface, realized P15 covers; also read the [separate price counterexample](three_fronts_20260924/P15_PRICE_BOUNDARY.md) |
 | Transformed-price budget, 24 September 2026 | [Proof](price_budget_20260924/PROOF.md) | A valid restricted extension, exact local-budget checker, and the unchanged 816-label application |
 | Full transformed-price range, 24 September 2026 | [Proof](full_price_20260924/PROOF.md) | All independent probabilities for demands at least 2; exact optimal uniform factor and the same palette; earlier counterexample remains valid at demand one |
+| Fixed-remote RN window, 24 September 2026 | [Proof](remote_window_20260924/PROOF.md) | Cubic mean count on a fixed spatial exclusion and between-pin height window |
+| Downstream hard gate, 25 September 2026 | [Package overview](downstream_gate_20260925/README.md) | Fail-closed promotion / reverse-impact integrity control for main #90/#86; scientific effect NONE |
 
 These directories are source snapshots, not competing current-status registers. Consult [main #67](https://github.com/d6g8k5htny-coder/main/issues/67) and its successors for reviews and corrections. The RN numerical integral, original 24-jet certificates, and unrestricted P15 problems are not closed by these entries.

frontiers/downstream_gate_20260925/D0_CI_UNBLOCK.md

@@ -0,0 +1,42 @@
+# D0 CI unblock — main PR87 packet allowlist
+
+**Object:** D0-PACKET-ALLOWLIST-FIX-20260925-v1  
+**Author:** Cursor. **Scientific effect: NONE.**
+
+## Diagnosis
+
+[main PR87](https://github.com/d6g8k5htny-coder/main/pull/87) (`chatgpt/downstream-crosswalk-20260925`) adds `docs/math_status/DOWNSTREAM_CROSSWALK_20260925.md`. Hardening `ci` / `math_status_check.py` fails with:
+
+```text
+PROBLEM packet: unexpected files ['DOWNSTREAM_CROSSWALK_20260925.md']
+```
+
+Observed on Actions runs `36135447237` and `36135443158`. Navigation and withdrawal-governance checks already SUCCEEDED.
+
+Root cause: `EXPECTED_NAMES` / `TRANSCRIPTION_NAMES` in `tools/math_status_check.py` and `PACKET.json` transcriptions do not yet list the new crosswalk file.
+
+Secondary hazard: the prose line `None sets D3 lemma_closed=true.` matches `ASSIGN_TRUE` and would fail after allowlisting; rephrase to `None promotes D3 lemma_closed.`
+
+## Exact fix (ready to apply on main)
+
+Portable patch: [`patches/d0_packet_allowlist.patch`](patches/d0_packet_allowlist.patch)
+
+Applies on top of PR87 head `3d4969ab40d36a38ebf6ea1c3608035a706a7576`:
+
+1. Register `DOWNSTREAM_CROSSWALK_20260925.md` in `TRANSCRIPTION_NAMES` and `NOTE_PHRASES`.
+2. Pin sha256 `d53b286029d034576245406152c188fe2470ccb089e51f73e9087e34bf5a0e10` (4804 bytes) in `PACKET.json`.
+3. Rephrase the `lemma_closed=true` negation as above.
+4. Docstring: six → seven transcribed bodies.
+
+Local verification after apply: `python tools/math_status_check.py` → `problems=0`, `lemma_closed=false`.
+
+## Why this is filed in Math-
+
+Cursor Cloud Agent has **Contents:Write on Math-** but received **403 pushing to `d6g8k5htny-coder/main`**. The fix was prepared and locally verified in a main worktree; it could not be published on that forge tip from this session. This note + patch are the source-bound D0/D7 unblock recipe for an agent/owner with main write access (or for merging into PR87).
+
+## Non-claims
+
+- Does not merge PR87.
+- Does not flip any scientific Boolean.
+- Does not edit Math- PR7.
+- Does not touch the #91 Drive vault.