Code Review and Readiness

Code review and readiness

Current publication review: 1.11.0-alpha.6 / documentation and media discovery

The alpha.6 prerelease publishes the current documentation/media-discovery changes as an exact template version. It adds no runtime capability or new schema. GitHub Actions remains disabled; release checks are run locally.

GitHub release 423158780 was published 2026-10-25 05:05:53 UTC (24 October, 22:05:53 PDT), with draft=false, prerelease=true, and immutable=true. Reviewed publication source is f942daf38fc7f1661910912dbcf07faaec5ad7aa; main, Alpha1, and the annotated v1.11.0-alpha.6 tag resolved to it at publication. All six public assets were downloaded without authentication and matched their reviewed byte counts and SHA-256 digests. Exact identities are in Versions and changelog and the public review.

The same complete repository suite was run independently under both interpreters on macOS arm64:

Python	Tests	Passed	Skipped	Failures / errors	Seconds
3.9.6	302	297	5	0 / 0	337.177
3.14.5	302	297	5	0 / 0	273.644

All five skips in each run require a case-sensitive filesystem unavailable on this test volume. These are repeated executions of the same suite, not distinct sets of tests. Ten publication-security tests passed separately on Python 3.12.14 with ReportLab 4.4.9 and Pillow 12.3.0. Those optional authoring libraries add no template runtime dependency. Windows and Linux were not freshly run for this release.

Strict template/package validation covers 237 portable files / 150 Markdown documents. Folder ZIP and wheel matched every manifest file, with no extra project files and zero package runtime dependencies. The 27 disposable local package/lifecycle commands completed with expected outcomes: fresh source and separate-wheel-environment preview/install/status/validation, logged removal and byte-exact restore of 478 files, and alpha.5-to-alpha.6 selection. That upgrade preserved 239 checked outer files, all 237 older cached-default files, and custom context, memory, ROOT, tasks, DATA, and notes. A pinned update refused without mutation. All eight stock bootstrap files were assessed and required no replacement; bootstrap refresh was a no-op.

The exact codeload tag ZIP is 18,337,363 bytes, SHA-256 a50f0566c07f5a0c51b14a38bf6e3eb38b3c61d2953352304b186fa4fca0b464. All 276 tracked source files matched reviewed Git blobs and the release stage. Five remote-pin commands completed with expected outcomes: preview, explicit-pin install, status, offline validation, and a wrong-pin refusal. The selected snapshot matched all 237 files with a manual current-version pin. The wrong pin returned exit 2 before creating .SYSTEMX, preserving the owner marker. The remote codeload pin is different from the folder-only ZIP checksum.

These checks establish the bounded local/package/publication results above. They do not establish live model/provider execution, cloud deployment, measured token/cost/time savings, or downstream production readiness. Earlier results below remain attributed to the releases that ran them.

The introduction retains media edition 1.0 and its alpha.5 feature reference: ten animated scenes/MP4 chapters, nine YouTube chapters, a 1080p film, and a separate 4K poster. White paper edition 2.0 retains its alpha.4 examined source and alpha.5 publication release. The four-part raster field guide retains its alpha.3 media edition. Publication does not establish external CLI execution, provider writes, downstream production readiness, or measured token/cost/time savings.

Historical publication review: 1.11.0-alpha.5 / paper edition 2.0

The published immutable prerelease was read back through GitHub's API: release 418402174, published 2026-10-20 13:01:32 UTC, draft=false, prerelease=true, immutable=true. Its reviewed publication source is 85c9d6b78b9143b7ad928cc2b26050d1194921fa. All five assets were downloaded while draft and matched retained candidate bytes; live API sizes/digests were verified after publication. At publication, main, Alpha1, and annotated v1.11.0-alpha.5 resolved to that source. GitHub Actions remained disabled.

This is a publication/manual update over the alpha.4 runtime, examined source e086ba15a1e70c6b585b8f12a703defc5c984b2c. Publication source: 85c9d6b78b9143b7ad928cc2b26050d1194921fa. The template version, paper edition 2.0, and unchanged alpha.3 four-part 4K media edition have different identities.

The final PDF has 24 pages, 4,728 words, four original figures, and 15 references, with selectable text, embedded Times New Roman, a professional MLA-style layout, linked citations, and branded headers/footers. Every rendered page passed publication review. Navigation inspection found 119 external HTTPS links, 104 valid internal links, and 15 outline entries. It is not a tagged PDF/UA export, independent peer review, or production certification. PDF: 1,130,293 bytes, SHA-256 def4d23341e1aabc6908cb02a330537af549f992b8fbe23151928ea6f8383771.

The source-grounded walkthrough completed 30 local CLI commands on both Python 3.9 and 3.14. Its synthetic local fixtures test task authority, acceptance refusals, many-to-many external-item aggregation, stale member fingerprints, known-bad evidence, and future-event rules. It starts no models, contacts no provider, creates no PR, and makes no deployment. This supports bounded protocol behavior rather than real external readback or comparative productivity. The paper's input-only accounting is hypothetical, not measured token, cost, or time savings.

New publication checks include 35 system runner tests on Python 3.9 and the same 35 on Python 3.14, and 10 publication-security tests on Python 3.12.14. These counts overlap; they are not a new full repository suite. Package/adoption parity covers 237 portable files / 150 Markdown documents. A fresh installed wheel passed first-run preview/apply and validation with only paper edition 2.0 in the new distribution/selected defaults. An alpha.4 project updated through preview/apply and validation while 231 checked outer files retained exact bytes, including the three old edition 1.2 paper files. All 228 files in the older defaults snapshot remained unchanged. Owner task, memory, rules, and local payload content was preserved. The selected new snapshot matched all 237 manifest files. These are distinct trees; the three old paper files are already included in the 231 count.

The complete alpha.4 suite and its targeted Python 3.9 results remain below, clearly attributed to the implementation release that ran them. Archived paper edition 1.2 remains available at v1.11.0-alpha.4; only edition 2.0 is active in the current paper folder. The four-page raster field guide is unchanged.

The exact codeload pin is 14e911de34ddc9e4be6151d474f669a50b732ae434ac51ee3458514295b769b2 (17,948,748 bytes); all 272 tracked source files matched the downloaded tag. Exact-pin remote install/validation passed. A wrong pin returned exit 2 before creating .SYSTEMX, preserving the owner marker. The five release asset sizes/hashes are retained in Versions and changelog and the public review.

Paper, metadata, example and archive · Format and draft-build instructions · Exact release evidence.

Historical generic-template cleanup review: 1.11.0-alpha.4

The published immutable prerelease was independently read back through GitHub's API: release 418362343, published 2026-10-20 12:20:00 UTC, draft=false, prerelease=true, immutable=true. Its target is reviewed source e086ba15a1e70c6b585b8f12a703defc5c984b2c.

This release removes the application-specific SDK/reference bundle and its integration page, advertising, navigation, contracts, and fixtures from the generic package/manual. Host applications own their service SDKs, domain schemas, data/UI, and adapters. SYSTEMX retains generic scoped coordination, planning, evidence, and the documented local/external-CLI toolchain.

Strict distribution validation passed on Python 3.9 and 3.14 for 228 portable files / 150 Markdown documents. The complete manager and system runner modules passed 96 Python 3.9 tests (61 manager, 35 runner), without skips; this is targeted module coverage rather than a second repository-wide full-suite run. The complete Python 3.14 suite ran 302 tests in 353.070 seconds: 297 passed and five were skipped, with no failures/errors. All five skips require a case-sensitive filesystem unavailable on the macOS test volume. The 96-method Python 3.9 manager/runner result is targeted coverage, not a second full-repository suite; overlapping counts must not be added.

The folder ZIP and wheel matched all 228 source files byte-for-byte. The fresh installed-wheel first-run preview/apply and validation contain no retired integration bundle. Package metadata is MIT-only with zero runtime dependencies. The current distribution text, wheel entry names, and wheel content were checked for removed-provider/host references.

An installed alpha.3 project updated to alpha.4 through preview/apply and validation. Its selected new 228-file defaults cache omits the bundle and test module, while 36 checked outer files retain their exact bytes: 31 retired reference files, including owner/custom content, plus five canonical/payload files. All 259 files in the older cached snapshot also remain unchanged. Every current media file matches the alpha.3 image/document bytes.

The current four-page graphic bytes retain 1.11.0-alpha.3 media edition and alpha.2 feature baseline. Historical release counts/checksums remain below with generic descriptions of removed host-integration history. Additive updates preserve existing adopted local files; removal from the public template is not permission to delete user records or rewrite history. Template checks do not establish an application's live service acceptance or production readiness.

Reviewed source: e086ba15a1e70c6b585b8f12a703defc5c984b2c. At publication, main, Alpha1, and annotated v1.11.0-alpha.4 resolved to the reviewed source. The source worktree was clean with local/upstream pointers matching; GitHub Actions remained disabled. All four assets were downloaded while the release was draft and matched retained candidate bytes. Independent live API readback confirmed release 418362343, draft=false, prerelease=true, immutable=true, publication 2026-10-20 12:20:00 UTC, and all four exact asset sizes/digests.

The exact-pin remote installation and validation passed with no retired bundle in the fresh outer records or selected defaults cache. A wrong pin returned exit 2 before .SYSTEMX creation, preserving the existing owner marker. The repository remains a public template; its description and topics describe generic project coordination rather than an application/provider integration.

Stack boundary · Upgrade preservation · Versions and changelog.

Historical ROOT infographic/media review: 1.11.0-alpha.3

The published immutable prerelease was independently read back through GitHub's API: release 417770466, published 2026-10-19 21:24:51 UTC, draft=false, prerelease=true, immutable=true. Its target is reviewed source e9428deef7175737e2a0151d8512d8e60d26fa65.

This media/manual update refreshes the four-page field guide to include ROOT internal rules/planning, SX-RULES.md, SX-MILESTONE.md, canonical task state, and observed harness/optional-bot TODO reconciliation. Agent 0 owns acceptance and authorized reconciliation with external readback; ROOT compares retained local observations and supplies no continuous connector. All four existing image filenames receive current release artwork; historical versions remain at immutable tags, and additive installations retain older local file bytes.

Strict template/distribution validation passed on Python 3.9 and 3.14 for 259 portable files / 163 Markdown documents. An independent focused ROOT suite passed 25 tests. The folder ZIP and wheel matched all 259 portable source files byte-for-byte. Fresh installed-wheel first-run preview/apply and validation passed.

An installed alpha.2 project upgraded to alpha.3 through preview/apply and validation. Ten checked existing files retained exact bytes: four earlier local graphics, an owner file, a DATA payload, ROOT PLAN/MIRRORS, TASKS, and project memory. The selected new defaults cache contained all four refreshed image bytes. Additive update preserves old root image files; select the new release snapshot or versioned public links to view the refreshed edition.

Runtime behavior remains the alpha.2 contract below. The earlier 306-test full Python 3.14 run and 77-method focused Python 3.9 run remain attributed to that source; a media update does not turn them into new full-suite coverage. The illustrations establish neither external tool execution nor cloud/PR/live acceptance, native platform certification, or a new Codex Security plugin pass.

Reviewed source: e9428deef7175737e2a0151d8512d8e60d26fa65. At publication, main, Alpha1, and the annotated v1.11.0-alpha.3 tag resolved to the reviewed source. The source worktree was clean with local and upstream pointers matching; GitHub Actions remained disabled. All four draft assets were downloaded and compared byte-for-byte before publication. The live release API confirmed 417770466, draft=false, prerelease=true, immutable=true, published 2026-10-19 21:24:51 UTC, and all four exact asset sizes/digests.

The installed wheel passed exact-SHA-pinned remote first-run and validation. A wrong all-zero pin returned exit 2 before .SYSTEMX creation and preserved the existing owner marker. The exact codeload ZIP is a separate byte stream from the folder-only release asset; use its digest for --archive-sha256.

Four-part field guide · ROOT manual · Versions and changelog.

Historical ROOT bounds and coordination-path review: 1.11.0-alpha.2

The published immutable prerelease was independently read back through GitHub's API: release 417384748, published 2026-10-19 14:23:57 UTC, draft=false, prerelease=true, immutable=true. All four retained downloads matched the uploaded asset bytes, and published API sizes/digests match the release table.

The patch measures ROOT packet/reply bounds against the complete rendered UTF-8 output and preflights mutation responses before writes. Agent Y's strict coordination guard includes ROOT/AUTOBOT for new preparation and selected execution/readiness while preserving original valid historical scopes for inspection. Existing non-ROOT path/secret/containment checks remain enforced. The public/manual review fixes first-use milestone/CWD prerequisites, matching preview/apply kinds, and planning vocabulary without changing graphic editions.

Final Python 3.9 focused verification passed 77 methods, without skips or timeouts: ROOT 25, runner 35, workspaces 13, and four Agent Y regressions. Strict validation covers 259 portable files / 163 Markdown documents. The full Python 3.14 suite completed 306 tests in 335.899 seconds: 301 passed and five case-sensitive-filesystem-only tests were skipped, with no failures or errors. The 77-method Python 3.9 run was targeted coverage, not a second full-suite invocation. Do not add overlapping focused/full counts.

Exact-package checks verified all 259 portable bytes. An installed alpha.1 project upgraded to alpha.2 while retaining adopted ROOT guides/requests/tasks, child ROOT records, and local DATA. The selected new runtime rejects oversized output without persisting a candidate even when older root files remain. Logged uninstall/restore preserved the observed 820-file synthetic inventory byte-for-byte.

Codex CLI 0.160.1 help and Copilot CLI 1.0.61 version/help, plus official GitHub CLI flags, were inspected read-only. These checks launched no real model/provider run and establish neither provider/PR/cloud acceptance nor native Windows/Linux end-to-end coverage. No new Codex Security plugin scan pass is claimed.

Published main, Alpha1, and the annotated tag name reviewed source 22d4a78f628443244b537c6688bb0b5bf0b3e059. The source worktree is clean and local/upstream pointers match. The public repository remains a template with GitHub Actions disabled. The correct explicit codeload SHA-256 pin passed remote first-run installation, recorded its pin receipt, and passed validation. An incorrect all-zero pin returned exit status 2 before creating .SYSTEMX. The published immutable prerelease is release 417384748, published 2026-10-19 14:23:57 UTC, with draft=false, prerelease=true, immutable=true. All four uploaded assets were downloaded and compared byte-for-byte, and their published API digests/sizes match the retained artifacts.

Versions and changelog · ROOT manual · Agent Y.

Historical ROOT planning/runtime review: 1.11.0-alpha.1

The published immutable prerelease names reviewed source cc8a43e964716ed7277745c267113f0de6372a70. This optional extension adds internal coordination rules, flexible planning nodes, originating/delegated request references, bounded tool cards/packets, and explicit harness/bot observations. It retains canonical task acceptance, existing Agent X/Y/Z lifecycles, and on-demand AGENTA routing. It adds no provider, cloud connector, background checks, or automatic review/acceptance loop.

Focused verification passed 21 ROOT tests on Python 3.9 and 3.14, plus 35 system runner and 13 project-workspace tests on Python 3.9. Strict template checks cover 259 portable files / 163 Markdown documents. The wheel matched all 259 portable source files byte-for-byte. Package first-run checks cover independent root/child initialization and validation; logged uninstall/restore retains the observed 558-file inventory byte-for-byte. The complete Python 3.14 suite ran 299 tests in 360.165 seconds: 294 passed, five case-sensitive-filesystem-only tests were skipped, and there were no errors or failures. The Python 3.9 run was a focused set of 69 passing tests (21 ROOT, 35 runner, 13 workspaces), with no skips; it was not a second full-suite run. The full-suite total already includes the ROOT tests and must not be added to the focused counts.

main, Alpha1, and the annotated tag resolve to the reviewed source. GitHub API readback confirms published prerelease 417217113, draft=false, prerelease=true, immutable=true, published at 2026-10-19 11:33:48 UTC, with all four asset sizes/digests matching retained downloaded bytes. A correct explicit codeload SHA-256 pin passed remote first-run, recorded the pin, and passed ROOT validation; a wrong pin was refused with exit 2 before .SYSTEMX creation. The 17,151,916-byte public codeload ZIP matched all 292 tracked source files; it is distinct from the folder-only distribution ZIP and wheel. The repository remains public and a template; GitHub Actions remain disabled. Exact archive and asset digests are listed in Versions and changelog.

Read the ROOT manual and Versions and changelog for exact release evidence. The 1.10 media edition and historical checks below retain their original scope. Local observations and passing template checks do not establish actual harness, cloud/provider, external PR, native Windows/Linux, or downstream production acceptance. No new Codex Security plugin scan result is claimed.

Historical media/manual review: 1.10.0-alpha.3

This is a media and documentation update for the alpha.2 feature contract; runtime behavior is unchanged. It adds a four-page 4K field guide: complete overview, Agent 0/X/Y/Z and CLI coordination, AGENTA with the ten exact AUTOBOT specialists, and DATA with named .SYSTEMXP records. The README, portable media index, text companion, prompts, provenance metadata and wiki connect all four pages. The older full-system JPEG moves to an immutable historical link in fresh distributions; additive updates preserve older local posters and user files.

All four exports are 3840 × 2160 JPEGs from 1672 × 941 native imagegen outputs, resized with macOS sips, JPEG quality 85. Upscaling adds canvas pixels, not native detail. Visual review checked the ten specialists with Alpha outside the slot count, all eight DATA folders, actual Agent Y/Agent Z record paths, explicit project scopes, and the on-demand reuse and acceptance boundaries.

Template/distribution validation passed for 249 portable files / 159 Markdown documents on Python 3.9 and 3.14. Six targeted tests passed on each interpreter; this patch did not rerun the full 272-method runtime baseline. The wheel's 249 portable files matched reviewed source bytes. Local lifecycle checks covered fresh installation, alpha.2 upgrade, addition of all four new pages, preservation of the older local poster, local DATA and context, and uninstall/restore preservation.

The alpha.3 public review and Versions and changelog record publication identity, artifact sizes/digests and the remote-install checks. Independent API readback confirms published prerelease, immutable=true, release ID 417180237, and the four asset sizes/digests. Correct/wrong raw tag ZIP pin installation checks pass; all four pages install. The retained alpha.2 results below are the broader runtime baseline; neither these media checks nor the diagrams establish live external provider/PR/cloud acceptance or production certification.

Four-part field guide · Visual Guide · Versions and changelog.

Historical Tier 2/detail and storage review: 1.10.0-alpha.2

This release adds the ten-slot AUTOBOT detail panel, separate Agent Alpha activation, bounded selected context, and deduplicated request/completion records. The local DATA scaffold includes exact tracked policy/READMEs while ordinary payloads are ignored by default; audit checks synthetic ignore probes and tracked index payloads without recursive payload reads. Those probes do not exhaustively verify actual untracked files and can miss custom per-file unignore exceptions; named export paths need separate git check-ignore review. Root/child ownership and lower-authority data boundaries are explicit in the source and manuals.

The immutable alpha.2 release names reviewed source fbc820f69833b4890b8dad01dfc02a49ecd77409, matching main and Alpha1 at publication. Final template validation passed for 244 portable files / 157 Markdown documents on Python 3.9 and 3.14. Exact ZIP/wheel checks matched all 244 source files. Integration exercised separate root/child request and completion reuse, additive alpha.6 upgrade, and uninstall/restore while preserving payloads, request records, and host files. A correct raw tag ZIP pin passed preview/apply and a wrong pin was refused. Public API readback confirmed published prerelease, immutable, and all four asset digests/sizes.

The public review records 272 distinct methods across full plus targeted runs, not a single final 272-test invocation. Python 3.14 passed a 270-test full suite (265 passes, five filesystem-case skips), then alpha.2 AUTOBOT 19 and system runner 35 checks. Python 3.9 passed a 267-test full suite (262 passes, five skips), then final DATA 15, manager 54, alpha.2 AUTOBOT 19, and runner 35 checks. The alpha.2 regressions include advertised 8 MiB request-ledger validation, unexpected DATA rejection before content inspection, strict 1.10 bundle requirements, and blank request-ledger integrity. Native Windows/Linux, live provider/PR/cloud, and host-application acceptance are separate unverified paths. See Versions and changelog for exact checksums and publication identity.

The earlier annotated v1.10.0-alpha.1 tag was not published as a release. Alpha.2 preserves that tag identity and corrects the request-ledger limit during validation plus strict-template DATA inventory ordering.

The architecture adds no model service, background process, recursive checks, cloud synchronization, authentication boundary, or production certification. Git-ignore does not remove history, hide tracked/force-added payloads, encrypt data, or constrain same-account tools. A historical completion retains its original proof and can become stale; a source fingerprint is not independent verification of a claim. Actual app checks, provider/PR access, transport, host permissions, and external acceptance remain the adopted project's responsibility. The earlier Codex Security Deep Scan could not start with this host's filesystem permission profile; no new plugin scan pass is implied by template tests. AUTOBOT manual · DATA manual · Versioned evidence.

Historical CLI/tooling and media review: 1.9.0-alpha.5

The alpha.5 release added direct installed systemx projects routing, the portable CLI/tools map with tools-guide/menu 18, reviewed unmanaged folder-copy launcher support, and invalid-managed-state refusal. The refreshed poster shows CLI entry points and exact named .SYSTEMXP scopes under one shared outer installation. Children retain separate project-owned records; app/IDE workers and Agent Y providers remain explicit execution choices.

The immutable alpha.5 release names reviewed source aba9620ac07c9fe27572fd37ba80df8d1deeded4. Its public review records 230 full-suite tests each on Python 3.14.5 and 3.9.6: 225 passes plus five expected case-sensitive-filesystem skips per run, no errors. The suite includes real PowerShell 7.5.5 launcher tests on macOS. Template checks covered 157 portable files / 98 Markdown documents, a 156-entry schema-2 manifest, and 12/12 blank seed hashes.

The exact ZIP/wheel matched all 157 committed portable file bytes. The offline wheel workflow passed lifecycle, Project-A/Project-B selection, role routing, uninstall/restore, and audit with user/host bytes retained. Alpha.4-to-alpha.5 first blocked outdated stock bootstrap files, then passed backed-up refresh and additive update while retaining the old map and adding the new one. An exact public-tag install passed the explicit raw codeload digest pin and child/tool-guide checks; a wrong digest was refused before installation.

The current 3840 × 2160 PNG/JPEG map is a sips export of 1672 × 941 native imagegen artwork; the JPEG uses quality 80. Independent tag readback matched its metadata hashes. Upscaling adds no native detail. The earlier four cards and paper keep their frozen baseline. The wiki audit covered 51 pages, 602 local references, 63 heading fragments, and 94 current source targets, with zero broken local targets. External URLs were not exhaustively validated. Published release asset sizes/digests matched the uploaded/downloaded bytes; release readback confirmed an immutable prerelease. The public review's pending-publication note preserves its generation-time state.

The current evidence does not establish a real provider, live GitHub PR write, native Windows/Linux end-to-end support, cloud transport, or downstream project production readiness. Agent Y native Windows execution remains disabled. Codex Security Deep Scan coverage remains unknown because the earlier scan could not start with this host's filesystem permission profile. Earlier test runs remain attributed to their historical source; this release does not turn those runs into new coverage. See CLI and Tools, Visual Guide, and Versions and changelog.

Historical documentation/media review: 1.9.0-alpha.4

This release refreshes the full-system poster and Agent Y instructions across the README, portable guides, and wiki. It preserves alpha.3’s runtime behavior and blank generic template records. The poster and focused diagrams now connect Agent 0/X/Y/Z, app-native workers, task-selected CLI worktrees, opt-in draft PRs, canonical records, .SYSTEMXP children, and versioned setup/recovery. Frozen edition 1.2 and the four environment cards remain historical Agent 0/X/Z media.

The immutable alpha.4 release names source 191dba6d18c130ac7ebc74a8e765542c08598917. Its public review record records template validation on Python 3.14.5/3.9.6, 156 portable files, 97 Markdown documents, 155 schema-2 manifest entries, and 12/12 blank seed hashes. The final focused Python 3.14 checks passed 31 runner + 48 manager tests with zero failures. Repository link/anchor validation covered 106 Markdown files, 655 local references, and 144 heading fragments with zero broken local references; external URLs were not exhaustively checked.

The exact ZIP and wheel contained 156 portable files each, matching committed source bytes. The fresh offline wheel workflow passed first-run, validation, status, uninstall, restore, and audit with user/host data preserved. The exact public tag ZIP passed explicit-digest remote install; the wrong digest was refused before target folder creation. Alpha.3-to-alpha.4 update preview/apply, validation, and audit preserved user context/custom files/older media/host bytes and needed no bootstrap refresh. All four published assets matched upload readback and API digests; publication readback confirmed a published immutable prerelease. The report’s prepublication note reflects when it was generated.

The alpha.4 PNG/JPEG are 3840 × 2160 exports of a 1672 × 941 imagegen edit; resizing adds no native detail. Three focused Mermaid/SVG/PNG sets were rendered and visually checked. Eight frozen baseline artifacts remain byte-identical to alpha.3. See Visual Guide for accessible text, hashes, and the complete four-role workflow.

The 223-test Python 3.9/3.14 runs below belong to alpha.3’s source. The current media/docs release does not claim a new full-suite run, native Linux/Windows runtime verification, live provider/PR writes, or a Codex Security scan pass. See Versions and changelog for exact source and assets.

Historical runtime review: 1.9.0-alpha.3

The alpha.3 review covers the distribution manifest's byte-exact binary/unknown file hashes, 1.9 Agent Y core-file requirement, unmanaged-target run refusal, Agent X closure chronology, and Agent Y ignored-file, Git privacy, remote-base, and provider-executable checks. These are template and controller checks; the external CLI still runs with its host account's permissions. Agent Y's allowed paths and PR gates are not an OS sandbox, and a GitHub base update is not atomic with gh pr ready. See Security and support and Agent Y before an authorized provider run.

The requested Codex Security Deep Scan could not start on this host because its managed filesystem permission profile was unavailable. No plugin scan result or coverage is claimed. A separate manual code review identified issues addressed in alpha.3 and the remaining provider/supply-chain boundaries; it is not a substitute for that plugin's scan. Template tests and offline Git simulations do not establish a real provider, GitHub write, cloud, or downstream project acceptance. Native Windows Agent Y execution remains disabled.

In a disposable local upgrade, the published alpha.2 folder archive was checksum-verified before installation. The alpha.3 manager refused selection until an explicit bootstrap-refresh preview and apply replaced the one changed stock manager.py; its byte-for-byte backup was verified. An additive update then selected alpha.3, retained both schema-1 and schema-2 releases, and added only the new white-paper erratum file at the project root. Offline validation, uninstall, restore, and final audit passed. A user-added file, edited context, and host-project file kept their original bytes. This local transition does not prove remote archive trust or another project's compatibility.

The immutable alpha.3 prerelease names source commit d43eab15397329512069f1e606eb4626d9a480d0, matching both main and Alpha1 at publication. Template validation covered 154 portable files, 153 manifest entries under schema 2, 96 Markdown documents, and 12/12 blank record hashes. The folder-only release ZIP and wheel each contained all 154 portable files with exact source bytes and no unexpected portable files. Ten local media-publication security tests passed. The 96-page white paper and four 3840 × 2160 infographic cards kept their original bytes; the separate erratum corrects one frozen citation.

Full macOS suites discovered 223 tests on Python 3.14.5 and 223 on Python 3.9.6, with zero failures and five expected case-sensitive-filesystem skips per run. Shell syntax, PowerShell parsing on macOS, and local documentation targets passed. A disposable exact wheel first-run/uninstall/restore lifecycle, an alpha.2-to-alpha.3 upgrade, and a digest-pinned public codeload install passed; a wrong digest was rejected before installation. All four downloaded release assets matched the uploaded bytes and their published SHA-256 digests. See Versions and changelog for the separate raw codeload ZIP digest and each asset hash. Native Linux/Windows runtime suites, real provider execution, and live Agent Y PR writes were not covered.

Previous alpha review: 1.9.0-alpha.2

The 1.9.0-alpha.2 source adds optional Agent Y CLI channels and guarded draft PR handoffs. The corrected source commit is 764af8df6fd5bb960776dc09567ebb00ef602bdc. Full local macOS suites ran 205 tests on Python 3.9.6 and 205 on Python 3.14.5, with zero failures and five case-sensitive-filesystem skips in each run. The disposable wheel install, uninstall, and restore roundtrip preserved custom user data and passed clean audits. A digest-pinned remote first-run preview and apply passed in a disposable project; an incorrect digest was rejected before install. See Versions and changelog for the exact tag, raw codeload ZIP digest, and published release-asset digests. GitHub reports this prerelease immutable; all four downloaded assets matched their uploaded copies byte for byte and SHA256SUMS verified them.

The earlier v1.9.0-alpha.1 tag is withdrawn: an exact pinned remote install found a distribution-manifest mismatch in tests/test_agent_y.py. Do not install it or treat its tag as the corrected release.

Offline controller checks do not establish real Codex/Copilot execution, GitHub write behavior, or production readiness. Those external paths were not tested end to end for this alpha. A partially successful remote action needs manual reconciliation; Agent Y will not retry it blindly. Native Windows agent-y run --apply fails before provider/worktree mutation because safe descendant-process termination remains unverified. The core Windows installer/record tools are separate from that optional execution path.

Previous release evidence: 1.8.9-alpha.1

The exact immutable prerelease and its public review record identify the source, artifacts, checks, and limitations. The annotated tag peels to f28c172496e28692d42024d1abdff9ece38ff02e. See Versions and changelog for separate release-asset checksums and the raw codeload tag ZIP digest.

Boundary	Recorded result and limit
Exact source and distribution	148 portable files and 92 Markdown documents validated. The ZIP and wheel each matched all 148 source files byte for byte.
Local behavior	Each macOS run discovered 182 tests on Python 3.9.6 and 3.14.5: 177 passed and five case-sensitive-filesystem tests skipped. Ten optional media-publication tests passed.
Package and migration	A disposable wheel install, offline validation, uninstall/restore, and a 1.8.8-to-1.8.9 upgrade preserved custom project files. A digest-pinned remote install succeeded and a wrong digest was rejected before installation.
Startup notice	One live opt-in check returned successfully in 0.594 seconds, retained the selected 1.8.9 defaults, and saved check metadata. The code limits the caller to a five-second wait; a timed-out metadata request may continue on its daemon thread.
Repository controls	Secret scanning, push protection, and Dependabot vulnerability alerts were enabled at publication. GitHub Actions and automatic Dependabot security updates were disabled. The main/Alpha1 ruleset blocks deletion and non-fast-forward pushes only. These settings do not transfer to generated repositories.
Still open	Native Linux and Windows runtime tests were not run. The Codex Security Deep Scan did not start because the host lacked its required managed filesystem permission profile; no plugin scan result is claimed.

The attached review record was prepared before the release was made public, so its publication.status still says release pending. The live release is published and marked immutable. Template validation and source checks do not certify an adopting project's build, permissions, deployment, or live behavior. Record that project's real actors and sources in the Access Matrix and System MAP, and apply Evidence and acceptance at the actual target. An Agent Z score cannot grant authority or replace evidence.

Earlier release evidence: 1.8.8-alpha.1

The exact 1.8.8-alpha.1 release and its public review record identify the reviewed source, artifacts, checks, and limitations. See the 1.8.8 artifact record for the raw codeload ZIP digest and separate release-asset checksums. This evidence is about the template release; an adopting project's stack, access controls, build, deployment, and live operation need their own checks.

Boundary	Recorded result and limit
Exact source and distribution	The reviewed v1.8.8-alpha.1 tag peels to bef5e54bc237d0e0b0e5a3de40897bbfcfc8261f; 143 portable files and 87 Markdown documents were inventoried.
Source behavior	175 tests passed on local Python 3.9.6 and 3.14.5 runs, with five filesystem-dependent skips on each. These were not native Windows/Linux runs.
Artifact and installation	Attached folder ZIP and wheel checksums were read back; a disposable digest-pinned remote installation and wrong-digest rejection were checked.
Startup notice	The new opt-in path reads release metadata without selecting a new version. A live check encountered a GitHub API rate limit and left installed defaults intact; notice delivery depends on metadata availability.
Still open	Native Windows/Linux installation and exhaustive passive PDF/image review were not established for this release. No downstream production acceptance is implied.

For an adopted project, record its real source, environment, actor permissions, and evidence routes in the System MAP and Access Matrix. Review applicable boundaries under Evidence and acceptance. A blank template check or an Agent Z score cannot replace those observations.

Historical documentation patch: 1.8.1-alpha.1

Version 1.8.1-alpha.1 adds the complete-project guide, shared workspace guide, current role diagrams, and README/manual consistency corrections. It changes no runtime behavior, project-record schema, or review-policy questions. The release verification report on Versions and changelog records that release's artifact checks. The runtime review below remains tied to its original 1.8.0 source/tests.

Runtime review: 1.8.0-alpha.1

The 1.8.0-alpha.1 review covered the portable command runner, task/agent records, project routing, installer and update manager, path handling, uninstall/restore, version parsing, shell and PowerShell launchers, Python packaging, tests, and public documentation. It also added the standard Agent X and Agent Z roles. This is a maintainer-assisted code review with recorded validation, not independent certification or a claim that every defect is known.

Findings addressed

Area	Change and reason
Workspace selection	Reject repeated global targets and abbreviations, including through a preserved older launcher, so arguments cannot redirect an already selected workspace.
Record boundaries	Match root and child handling of symbolic links, junctions, and hard links; reject linked writable coordination records.
JSON integrity	Reject duplicate fields and non-finite numbers in manager and coordination/configuration inputs instead of accepting ambiguous values.
Distribution inputs	Validate malformed distribution and blank-record manifests with readable errors.
Child configuration	Return a controlled error for invalid child command configuration instead of an uncaught exception.
Growing histories	Remove repeated sibling scans from child guards; bound role records at 32 MiB and refuse oversized replacement before writing. Event views return a selected, limited result.
Library contract	Export the documented run function and retain explicit argument arrays and scope selection.
Interrupted setup	Record failed child-creation receipts on interruption. Standard-role setup also records completion or failure.
Role integration	Return scoped role results as structured JSON, reserve role IDs, and preserve existing project policy and memory.

Verification and reproducibility

The release includes 126 inventoried distribution files and 78 checked Markdown documents. Its local macOS run discovered 143 behavioral tests: 138 passed and five were skipped because the corresponding filesystem behavior was unavailable. The suite exercises command execution boundaries, paths, additive updates, coordination/acceptance, child isolation, lifecycle recovery, role activation, event replay, due tracking, all 100 rubric IDs, answer validation, score math, policy revisions, report integrity, and review deltas.

An isolated wheel installation verified the full packaged inventory. A project installed from the published 1.7.0-alpha.1 tag then upgraded to the new defaults: every pre-existing file except manager-owned installation state was preserved. The preserved older launcher successfully activated and used X/Z records in the root and a selected child. An unselected sibling remained unchanged. Repeated events and score requests reused their records without writes. A locally edited policy survived an update. Uninstall and restore preserved all recorded file hashes, child code, role histories, and an empty user directory; the external receipt recorded successful restoration.

See Versions and changelog and its machine-readable historical release verification report for exact source revisions, completed platform CI, download checksums, and publication verification. Tests use disposable fixtures; the distributed project/task registry stays blank and only Agent 0 is registered until an owner explicitly activates additional roles.

To repeat source checks:

python3 -I -B .SYSTEMX/scripts/release.py --check
python3 -I -B -m unittest discover -s .SYSTEMX/tests -v

Readiness limits

• The release remains alpha; use at your own risk. These results establish tested template behavior, not a downstream application's production readiness.
• Local file locks coordinate cooperating writers. They are not distributed locks, a database transaction system, or protection against hostile concurrent filesystem changes. Keep one owner for shared writes.
• Event ledgers are bounded local JSON records. They are not a high-throughput event bus; plan archival/partitioning and measure workload limits before accumulating long histories. No 10,000-event performance guarantee is made.
• Agent X tracks schedules and observations. It does not install a daemon, execute jobs, supply recurrence/DST handling, or authenticate external actors. The selected scheduler and harness own those functions.
• Agent Z fixes the questions and arithmetic. Humans or authorized models make the judgments. An evidence reference is not proof of authenticity, and a hash is not a signed attestation. Critical findings and acceptance criteria remain more important than the aggregate score.
• Activated X/Z registries require the new tools. Update the external library to receive manager fixes; preserved project bootstrap files are not rewritten. Downgrading defaults after activation requires reviewed record migration or a pre-activation backup. See Upgrading.
• No provider-specific deployment, cloud application, private project records, or credentials are required by the template. Connected projects still need their own runtime, access, security, delivery, and live verification.

Use the 100-question policy to document those distinctions for a selected project. Reuse unchanged evidence, record what changed, and let Agent 0/user choose the next bounded action or accept completion.