Remove redundant implementation JSON reports

Author: KSemenenko

.gitignore

@@ -679,3 +679,11 @@ output/playwright/
 *.plan.md
 *.brainstorm.md
 *.acceptance.md
+
+# Canonical documentation records only; intermediate reports belong in CI artifacts.
+/docs/implementation/*.json
+!/docs/implementation/documentation-coverage.json
+!/docs/implementation/mcaf-installation.json
+!/docs/implementation/sql-client-commands-postgresql18.json
+!/docs/implementation/sql-client-conformance.json
+!/docs/implementation/status.json

AGENTS.md

@@ -467,6 +467,7 @@ For changes outside existing owner authorization, obtain direction before changi
 - Every public KeyLoad page ends with "Developed by Managed Code" and a normal followed (dofollow) link to https://www.managed-code.com/.
 - The central KeyLoad K MUST use the canonical true SVG outside the bounded 3D raster buffer so it stays sharp on Retina displays. An SVG wrapper around a bitmap or a CanvasTexture logo does not satisfy this requirement (repeated owner correction 2026-10-03).
 ### Dislikes
+- Owner correction 2026-10-06: do not create or commit intermediate per-run, per-source-review, checkpoint or development JSON reports in `docs/implementation/`. Keep working reports outside the checkout and generated verification results in CI artifacts; use direct original GitHub links in durable feature/ADR records. Retain only canonical implementation status, documentation traceability, repository installation validation and versioned SQL inventories there. Remove redundant reports and their local links together without fabricating qualification.
 - Owner reiteration 2026-10-03 requires immediate cleanup: delete obsolete, unused, superseded, duplicate and abandoned code, legacy implementations, temporary artifacts and discarded implementation/evaluation plans in the same coherent change. Remove their obsolete tests, configuration, dependencies, routes and documentation references together; do not retain dead branches, compatibility shims or fallback implementations for an unspecified later cleanup. Preserve required behavior with the real replacement and its mapped acceptance tests, unrelated active work and persisted user data. Existing historical migration-debt records do not authorize retaining a replaced implementation.
 - Repeated permission questions for already authorized work.
 - Working `*.plan.md`, `*.brainstorm.md` and `*.acceptance.md` files MUST NOT be created or committed anywhere in this checkout. Remove existing copies and ignore these patterns without exceptions. Keep durable requirements, acceptance criteria and execution/verification contracts in the owning `docs/Features/` specification and required ADR; agent working notes stay outside the repository. This explicit owner correction on 2026-10-03 supersedes every root/local instruction requiring separate planning files without weakening feature, test, architecture or qualification requirements.

docs/ADR/ADR-005-canonical-keyspace-codec.md

@@ -100,7 +100,7 @@ restores prior validation while preserving valid durable keys; its malformed
 exception leaks remain explicitly unqualified. SDK/MCP/frontend changes N/A:
 unchanged database operations. Dependency repair N/A: the defect is KeyLoad-owned.
 
-The [2026-10-04 local receipt](../implementation/keycodec-crud-development-2026-10-04.json)
+The 2026-10-04 local receipt (report removed from repository)
 records14 new KeyCodec cases and4 original cases in unchanged full Aspire
 normal/scalar suites at2889/2889 each, recovery228/228 and1000 unique actual
 atomic process cuts. Literal fractional decimal goldens pass under invariant

docs/ADR/ADR-035-memory-performance.md

@@ -823,7 +823,7 @@ transport/error join stays with the lead to serialize contract and evidence edit
 
 REQ-STORAGE-014 / AC-RC-001..004 / TASK-REC-ADMIT-002 confines the repair to
 test infrastructure. The exact c486 Windows cancellation baseline is recorded in
-[the complete native receipt](../implementation/runtime-qualification-37060131271.json).
+[the complete native receipt](https://github.com/managedcode/KeyLoad/actions/runs/37060131271).
 Ordered implementation: freeze acceptance; one bounded four-slot shared storage
 trial owner; acquire before each original15s/20s deadline; retain every real
 CrashHost stage and recovered-cut/retry/metadata assertion; release after cleanup;

docs/ADR/ADR-040-static-site-threejs-evidence.md

@@ -438,7 +438,7 @@ publication and complete acceptance evidence exist.
 
 ## Scoped implementation evidence, 2026-10-02
 
-REQ/AC-BC-011–018/024/025/027/028 are implemented and qualified at exact H `6a82c86d0113270335368bbfbff0080ea1c1800a`, with strongest TASK-SITE-REVIEW-007 documentation/evidence join COMPLETE. [Canonical status and immutable receipts](../implementation/site-design.json) bind complete GitHub validation37011817610, automatic push37013009381 and workflow_run37014111869 to118+70 non-skipped tests, unchanged native and JavaScript source/critical gates, actual predeploy tuple, original12/final9 raw files and exact35-file Pages output. All previously failed attempts remain retained.
+REQ/AC-BC-011–018/024/025/027/028 are implemented and qualified at exact H `6a82c86d0113270335368bbfbff0080ea1c1800a`, with strongest TASK-SITE-REVIEW-007 documentation/evidence join COMPLETE. Canonical status and immutable receipts (report removed from repository) bind complete GitHub validation37011817610, automatic push37013009381 and workflow_run37014111869 to118+70 non-skipped tests, unchanged native and JavaScript source/critical gates, actual predeploy tuple, original12/final9 raw files and exact35-file Pages output. All previously failed attempts remain retained.
 
 Both automatic publications succeeded at https://www.keyload.cloud/; current provider evidence confirms the authorized www/HTTPS update and live apex301 redirect. Live HTML/publication/catalog/raw-report bytes match qualified output. Manual desktop/mobile/WebGPU observations supplement the real GitHub Chrome qualification and preserve the narrower physical-device-loss exception. Schema3/advanced benchmarks, database/RF3/endurance/power-loss qualification and production readiness are separate; their status is not changed by this ADR closure. DNS was not modified.
 

docs/ADR/ADR-050-timeseries-timescale-comparison.md

@@ -61,7 +61,7 @@ flowchart LR
 
 ## Implementation contract
 
-1. Root adds `ManagedCode.TimeSeries` to `Directory.Packages.props`, adds its reference only to `benchmarks/KeyLoad.Comparisons/KeyLoad.Comparisons.csproj`, and owns shared comparison profile/report registration. The original pin was published `10.0.0`; the temporal repair is published `10.0.2` and the current shared pin advances to `10.0.3` after the owning summer allocation repair and independently verified [release/feed receipt](../implementation/timeseries-dependency-10.0.3.json). Historical reports keep their original package identity; new exact-source comparison and consumer regression execution is mandatory.
+1. Root adds `ManagedCode.TimeSeries` to `Directory.Packages.props`, adds its reference only to `benchmarks/KeyLoad.Comparisons/KeyLoad.Comparisons.csproj`, and owns shared comparison profile/report registration. The original pin was published `10.0.0`; the temporal repair is published `10.0.2` and the current shared pin advances to `10.0.3` after the owning summer allocation repair and independently verified release/feed receipt (report removed from repository). Historical reports keep their original package identity; new exact-source comparison and consumer regression execution is mandatory.
 2. Root owns `src/KeyLoad.AppHost/Features/BenchmarkComparisons/Resources/BenchmarkResources.cs` and `src/KeyLoad.AppHost/Features/BenchmarkComparisons/Resources/TimeSeriesBenchmarkResources.cs`. Register the Timescale resource only for the `timeseries` benchmark profile; use `timescale/timescaledb:2.30.2-pg18@sha256:e72689191e1c977892c53d6f2c344dbc4a9657a867dc8cc1899229f9d3672b2e`, parameterized Aspire connections, and readiness ordering. The container is ephemeral and has no cross-run data volume; the report claims persistence only for committed rows during that container's lifetime. Do not start it in ordinary RF3 product fixtures.
 3. Root owns deterministic profile contracts/data/report and `benchmarks/KeyLoad.ComparisonHost/Features/BenchmarkComparisons/` composition. The shared oracle checks UTC buckets, exact values, order, range edges and duplicate identity. Timings separate persistent append/read/aggregate work from in-memory aggregation; retain every failure.
 4. A bounded implementation task owns new Timescale Npgsql and ManagedCode.TimeSeries target files under `benchmarks/KeyLoad.Comparisons/Features/BenchmarkComparisons/TimeSeries/`. Use parameterized SQL, isolated per-run namespace/owner marker, hypertable-aware identity and cleanup only after positive ownership acknowledgement.

docs/ADR/ADR-052-timeseries-bounded-aggregates.md

@@ -21,14 +21,14 @@ Temporal rounding defects were subsequently repaired in the owning repository;
 published10.0.2 source2c9118b6fbd45a5170682d345d66cba4ad6e8f2d passed1104/1104
 native tests and90% coverage in each module. The lead independently verified the
 release/tag, actual package contents, NuGet signatures and duplicate-publication
-fence in the [delivery receipt](../implementation/timeseries-dependency-10.0.2.json).
+fence in the delivery receipt (report removed from repository).
 The following summer allocation patch is published10.0.3 at source
 de44e91ecf38779b61fe79a51153ce7b7b4db0ea. Its owning Release run37077860613
 passes1107/1107 tests, Core687/725 and Orleans205/212 covered lines. The lead
 independently verifies the annotated tag, actual GitHub assets and both signed
 NuGet payloads before the new central pin. Four native repeated normal/scalar
 profiles complete16 cases each, with20 result iterations per case; the
-[10.0.3 receipt](../implementation/timeseries-dependency-10.0.3.json) keeps
+10.0.3 receipt (report removed from repository) keeps
 source, package and measurement proof separate. The private static generic-state
 callbacks preserve the public accumulator contract and add no consumer
 workaround. The KeyLoad10.0.3 consumer change requires a new exact-SHA GitHub

docs/ADR/ADR-057-orleans-atomic-wal.md

@@ -62,6 +62,6 @@ Run37077856823 at6ad4741a7 disproved the authored1024-byte size assertion and ov
 
 ## Current mandatory-gate evidence
 
-The [native receipt](../implementation/atomic-wal-qualification-37084177131.json) joins authenticated run37084177131, exact source cf630751e0f24e4d8183e55510add9c7207e377f, original ZIP digests and terminal successful jobs. Full Release restore/build, formatter/governance and118 analyzer cases pass; units1407/1407 in normal and scalar modes, recovery136/136 and genuine RF3 SDK/MCP63/63 pass with no skips. All66 WAL cases pass in each mode; their19 source methods and native expanded signatures match the earlier independently joined c10c source, with this storage/test slice unchanged. All1000 real process-kill receipts retain atomic cuts and values at every observed cut at or after journal flush.
+The native receipt (report removed from repository) joins authenticated run37084177131, exact source cf630751e0f24e4d8183e55510add9c7207e377f, original ZIP digests and terminal successful jobs. Full Release restore/build, formatter/governance and118 analyzer cases pass; units1407/1407 in normal and scalar modes, recovery136/136 and genuine RF3 SDK/MCP63/63 pass with no skips. All66 WAL cases pass in each mode; their19 source methods and native expanded signatures match the earlier independently joined c10c source, with this storage/test slice unchanged. All1000 real process-kill receipts retain atomic cuts and values at every observed cut at or after journal flush.
 
 This qualifies the specified WAL gates at that exact source SHA. The overall workflow still ran the separate isolated comparative cohort at receipt capture; it was not a full green workflow or an acceleration result. ADR status stays Accepted while numeric coverage, decoded-memory/work amplification and remaining malformed-envelope/current-header proof are open. Process-kill recovery does not establish power-loss durability, endurance or production readiness.

docs/ADR/ADR-058-orleans-coordinated-cache-memory.md

@@ -481,7 +481,7 @@ flowchart LR
 
 REQ-CACHE-002/003/007 / AC-CACHE-015. Exact native baseline, original ZIP
 digests,17 unit failure IDs and the distinct RF3 failure are retained in
-[run37093992229](../implementation/runtime-qualification-37093992229.json).
+run37093992229 (report removed from repository).
 This test-only repair preserves existing provider identity/copy/error contracts.
 
 1. Root freezes AC015 before delegated writes; the existing native failures

docs/ADR/ADR-059-isolated-intensive-timeseries.md

@@ -542,7 +542,7 @@ DDL and source compilation alone cannot authorize native performance claims.
 
 TS007F actual pinned-image feasibility passes at890047996, run37086903497/
 job111098989273,1/1 with no skips. The exact native evidence is
-[image receipt](../implementation/timeseries-image-qualification-37086903497.json).
+image receipt (report removed from repository).
 Observed image: Linux/amd64 Alpine3.23.6, PostgreSQL18.6, gosu at
 /usr/local/bin/gosu, postgres UID70 and actual writable PGDATA18/docker. Owned
 container removal and retained command hashes pass. This resolves the native