Preserve incomplete WSL retrieval evidence and enforce provenance checks

.github/osv-scanner-lockfiles.json

@@ -19,6 +19,9 @@
     {
       "path": "blueprints/convergence-practice/application-delivery/uv.lock"
     },
+    {
+      "path": "blueprints/convergence-practice/wsl-retrieval/package-lock.json"
+    },
     {
       "path": "blueprints/us-equities/adaptive-paper/requirements.txt"
     },
@@ -137,6 +140,10 @@
       "path": "blueprints/convergence-practice/application-delivery/pyproject.toml",
       "lockfile": "blueprints/convergence-practice/application-delivery/uv.lock"
     },
+    {
+      "path": "blueprints/convergence-practice/wsl-retrieval/package.json",
+      "lockfile": "blueprints/convergence-practice/wsl-retrieval/package-lock.json"
+    },
     {
       "path": "blueprints/gap-wave2-20260923/us-equities__portfolio-risk/requirements-libs.in",
       "lockfile": "blueprints/gap-wave2-20260923/us-equities__portfolio-risk/requirements-libs.lock"

blueprints/convergence-practice/layer-acceptance.md

@@ -15,6 +15,7 @@ and deployment remain distinct.
 | [WSL transport](wsl-transport-recovery/README.md) | Fresh reconnect found the same detached Dagu job running; one durable checkpoint, native completion without retry, original12-test oracle before/after | Controlled owned SSH-client termination; failed first harness attempt retained. No real network outage, host restart or general model-session claim. |
 | [WSL documents](wsl-document-ingestion/README.md) | Native Poppler26.09.0 build passed the unchanged3-page/24-cell/12-query protocol; all four output artifacts match Mac hashes | Task-only source build and verified dependency extraction. Designed corpus and exact lexical retrieval only. |
 | [WSL memory](wsl-memory-maintenance/README.md) |65 native checks and39 MCP calls; scoped restart, upstream HTTP backup and native restore, four owned processes exit0 | Fresh synthetic store with capture and embeddings off. Production service and native-client activation are separate. |
+| [Historical WSL retrieval reference](wsl-retrieval/README.md) | Retained receipts report 22 source and 70 QMD checks; offline consistency only | Incomplete invocation provenance: argv/cwd absent and raw logs unavailable. Historical blanket npm script suppression is unsupported. Initial URI-checker failure retained; no native E2E or adoption acceptance. |
 
 [WSL native quality tools](wsl-native-tools/README.md) add verified gitleaks8.30.1
 and Worktrunk0.78.0 with clean/secret controls, dirty-worktree refusal, exact pin

blueprints/convergence-practice/wsl-retrieval/README.md

@@ -0,0 +1,112 @@
+# Incomplete historical WSL retrieval reference
+
+The retained September 20, 2026 receipts are an **incomplete historical fixture
+reference**. They do not establish native E2E or adoption acceptance. Per-command
+`argv` and working directories were not retained, and the original private raw-log
+locations could not be verified. Selected outputs and hashes can be checked for
+consistency, but cannot independently identify the commands that produced them.
+The historical QMD installation also disabled all npm lifecycle scripts, which is
+not the supported fresh-host recipe. The current assessment therefore defers
+acceptance in [experiment.json](experiment.json).
+
+The receipts report these outcomes, without a new native run during offline review:
+
+| Retained attempt | Reported outcome | Recorded commands |
+| --- | --- | --- |
+| [Source attempt 1](source-receipt.json) | 22 checks passed | 6 |
+| [QMD attempt 1](qmd-attempt-1.json) | URI-checker assertion failed | 5 |
+| [QMD attempt 2](qmd-receipt.json) | 70 checks passed | 17 |
+
+The source fixture contains the frozen planner from repository revision
+`6f74bc503ccecaaf6ccebd53677b23aedab50921`. Its selected ripgrep **15.2.0** and
+ast-grep **0.45.3** results contain exact file lines, byte spans and source text.
+The QMD **2.8.3** fixture reports Node **24.21.0**, better-sqlite3 **13.0.3** and
+sqlite-vec **0.1.9**, with the retained [package lock](package-lock.json). It covers
+five positive and two negative queries, Unicode, a one-line `get`, process reopen,
+update and deletion in a three-document primary collection plus one decoy.
+The final recorded database has integrity `ok`, three active documents and zero
+vector rows. These are receipt contents, not independently recovered execution.
+
+These locally authored integration checks use synthetic documents and a frozen
+source file. Under the [acceptance evidence policy](../../../docs/acceptance-evidence-policy.md),
+they cannot substitute for unchanged upstream tests. No semantic RAG, production
+corpus, native client, service, inference, OS-confinement or token-saving acceptance
+is claimed. Output bounds are assertions after capture, not streaming limits.
+
+## Preserved history and the provenance gap
+
+The initial QMD checker compared an entire URI with a bare collection/path. The
+selected output included the expected body and `?index=wsl-retrieval-fixture`, but
+`qmd-positive-0_paths` failed. [run-initial.py.txt](run-initial.py.txt) preserves the
+source attempt and failed QMD attempt's runner. The corrected historical runner
+is now archived as [run-qmd-attempt-2.py.txt](run-qmd-attempt-2.py.txt), with its
+original SHA-256 `4cbcceac02158629ca78262aaa826c995c21bbe45fa83481f7c139f16a6c52d2`.
+The successful QMD receipt originally mapped `run.py` directly; the offline audit
+now resolves that historical name to the archived bytes. Neither receipt was
+rewritten to pretend it recorded today's file or missing invocation metadata.
+
+The [install receipt](install-receipt.json) still records the actual historical
+`npm ci --ignore-scripts --omit=optional` action. Its contents and digest are
+unchanged. The [inventory](install-inventory.json) reports no optional llama
+backend or model files, but does not qualify fresh native dependency setup.
+[pins.json](pins.json) and [source-review.json](source-review.json) preserve dated
+identities; release currency remains September 20, without a new latest-release
+claim from the September 23 offline completion or subsequent review.
+
+Original handoffs reported full stdout/stderr logs under private run directories.
+Their exact locations remain unavailable, and those logs have not been reopened.
+No command line or working directory has been reconstructed as an observation.
+The historical [verification.json](verification.json) preserves the earlier
+18-test offline check record; its original acceptance wording is superseded by
+this incomplete assessment. Failed attempts and original evidence remain intact.
+
+## Offline verification
+
+From the repository root:
+
+```sh
+python3 blueprints/convergence-practice/wsl-retrieval/audit.py
+python3 -m unittest tests.test_wsl_retrieval -v
+python3 scripts/validate_convergence.py \
+  blueprints/convergence-practice/wsl-retrieval/experiment.json --root . --json
+python3 scripts/validate.py
+```
+
+The [audit](audit.py) requires the exact frozen-input and check-name sets for all
+three receipts before comparing hashes and results. Mutation regressions reject
+omitted inputs and same-count replacement checks as well as incorrect spans,
+source bodies, URI scope, stale updates, erased failures and unsupported claims.
+A zero audit exit means retained facts are consistent; its result explicitly
+reports `native_acceptance_established: false`. It cannot repair missing evidence.
+
+## Future command capture and supported installation
+
+[run.py](run.py) is a future recording aid. It now retains each attempted command's
+sanitized argument vector and working directory before launch, preserves argument
+boundaries, and records failed launches and timeouts. Private path roots become
+scope markers such as `<RUN>` and `<NODE>`; original historical facts receive no
+fabricated fields. Mocked regressions exercise this recording without invoking
+native retrieval. Both previously executed runner versions remain archived.
+
+For a separately authorized fresh QMD trial, follow the maintained
+[QMD native recipe](../../../recipes/README.md#component-catalog-install-and-check) and
+[native token fixture guidance](../../../docs/native-token-ci.md). The supported
+installation control is:
+
+```sh
+NODE_LLAMA_CPP_SKIP_DOWNLOAD=true npm install --global \
+  --prefix "$NEW_OWNED_QMD_PREFIX" @tobilu/qmd@2.8.3
+```
+
+Required npm dependency lifecycle scripts remain enabled. Do not reuse the
+historical blanket `--ignore-scripts` command as a fresh-host recipe. The retained
+lock and optional-backend assertions describe the historical prefix; they do not
+prove compatibility of a newly installed prefix. Use the maintained native fixture
+for current supported installation acceptance, retaining its actual install and
+runtime commands and outputs. This review performs no install or native rerun.
+
+Acceptance requires recovered verifiable historical invocation evidence or a
+separately authorized reachable-host trial with supported installation and complete
+command capture. Preserve native accounts, model/effort settings, caching and
+compaction. No automatic history capture is introduced. Whole-task provider,
+parent/child/retry/cache usage remains unknown; no savings claim is made.