AI-Assisted Design Engineering Operating System v3.9.0

AI-Assisted Design Engineering Operating System v3.9.0

This release introduces three optional generic tools, extending v3.8.0 without automatic project adoption.

New capabilities

• runtime/scripts/context_read_guard.py: bounded contextual source delivery, integrity hashes and source manifests.
• runtime/scripts/check_action_claim.py: consistency checks for supplied tool-action evidence. Does not independently authenticate external tool execution.
• runtime/scripts/op_cycle.py: explicit, read-only cycle-focus packets for a selected Workstream. Does not grant mutation authority or independently reconstruct objectives.

Compatibility

• Canonical runtime package: 62 to 65 entries.
• 32 shared and 33 configured entries.
• 35 documented methodologies.
• Seven optional pilots remain optional.
• No mandatory bootstrap changes or automatic private-runtime migration.

Validation

Exact release source:
795cedc

Operator-executed local Windows QA:

• 9/9 official workflow-equivalent commands PASS
• 880 framework package checks PASS
• Current-state tests 8/8 PASS
• Symlink capability READY
• SKIPPED_TESTS=0
• Clean detached checkout

GitHub-hosted CI: NOT VERIFIED. This is not a hosted CI PASS.

ADR-0036 records a v3.9.0-only acceptance of this documented risk under V3_9_0_MANUAL_HOSTED_SUSPENSION_DISPOSITION.

These tests do not prove independent agent effectiveness, externally authenticated tool actions, or acceptance of optional methodology pilots.

Licensing

Software-like files: PolyForm Noncommercial 1.0.0.
Documentation: Creative Commons Attribution-NonCommercial 4.0 International.

Commercial rights remain reserved. These are noncommercial, non-OSI licenses.

Adoption and corrections

Publication does not modify private project runtimes, dependency pins, configured files, Workstream states or historical evidence.

Project adoption requires separate compatibility analysis, three-way reconciliation and QA.

If corrections are needed, preserve the original tag and release history, publish transparent errata, and use a separately reviewed corrective version.

Evidence

Final release-source QA and governance:
#28 (comment)

Framework v3.8.0 — controlled contextual reading

Optional GOV-CTX-001 adds explicit contextual reading, original source sections and hashes, activity/Workstream routing and strict route integrity. The full 54-section Master is preserved (version header only); the default 00–70 bootstrap, existing contracts, schemas/enums, permissions, licenses and pilot/proposal status remain intact. The existing comments-export procedure is classified in the exhaustive package index. No project activates automatically.

Released source: 0d0b8ffffa3a640d937e97c54062c517c5b88029; Git tree: ff755fb4c800b440fe12d1bf53319195cc3c92e7. PR #21 was integrated with exactly the tested tree. All 21 changed remote contents and the entire 200-file tree matched the tested payload/baseline; full public privacy/rights and preservation review passed.

Eight local workflow-equivalent steps passed on Windows/Python 3.12: package 830 checks; synchronization; methodology17; catalog35 without errors; catalog regressions23; onboarding7; current-state8 (one host-restricted symlink subcase skipped); contextual reader3 including14 negative cases and real Portuguese output under CP1252. Local fixtures and subprocess tests do not establish AI compliance or live product behavior.

Hosted QA: NOT RUN. Static QA was visibly manually disabled and remains suspended by explicit maintainer instruction. ADR-0034 governs only this exact3.8.0 local-validation/online-consolidation scope; local Windows PASS is not hosted Linux PASS or fresh ADR-0022 runner-zero evidence. No automatic reactivation or future waiver is introduced. ADR-0033 governs this optional method. ADR-0032 remains PROPOSED; distribution does not accept it or promote pilots.

Adopt only this immutable released source through three-way shared-file comparison and an explicit project decision. Preserve configured data/history/ownership/report cursors and rollback references. Legacy full bootstrap remains available. The reader never verifies consent or authorizes execution; emitted bytes do not prove token/cost/productivity savings. Independent fresh-chat trials belong to the actual project adoption.

Licensing remains PolyForm Noncommercial1.0.0 and CC BY-NC4.0 with commercial rights reserved; source-available for noncommercial use, not OSI Open Source. No proprietary content was imported.

Framework v3.7.0 — evidence-based validation and reusable design methods

This compatible minor package consolidates accepted validation policy, reusable design guidance and scoped evidence while preserving the v3.6 runtime architecture.

Added and improved

• Provider-neutral AI-assisted installation/configuration guidance, explicit immutable source selection and fictional onboarding/fresh-process regression tests.
• Accepted GOV-VAL-001 / ADR-0029: validation proportional to risk, reuse of still-valid evidence, protected mandatory criteria and closure limited to the evidenced deliverable.
• Generic evidence/delivery, Design System QA and methodology guidance, a bounded 35-topic catalog and traceable documentary transfer checkpoint.
• Optional method-selection guide, component-behavior template, fictional failure/recovery specification and derived decision brief, with pinned external attribution and reuse limits.
• Reproducible fictional QA tools, scoped experiment reports, a downloadable human-reading packet and native annotation/layout observations with documented caveats.
• Current-package, synchronization, methodology, catalog and onboarding checks; generic evidence-ingestion and optional local transcription guidance.

Compatibility and adoption

The package retains the v3.6 runtime architecture, startup intents and mandatory bootstrap order, inventory paths, runtime schemas/states, configured project values and compatibility aliases. Framework version, installed runtime version and schema versions remain independent. No populated project data is included and no installed runtime updates automatically.

Before adopting, verify the actual released immutable commit and follow the synchronization procedure. Compare the installed declared source, actual shared files and target. Reconcile local shared-file divergences, review new configured seeds, preserve unknown extensions/history and obtain project-specific migration approval. Keep a rollback source.

Maturity and known limits

GOV-VAL-001 is accepted operating policy. ADR-0024–0028 and ADR-0030 remain PROPOSED; their supplemental guidance and experimental material are distributed with that status. The seven optional pilots retain their classifications. A release does not accept proposed decisions or validate pilot effectiveness.

• Human comprehension, individual native annotation collapse/expanded occupancy and cross-file published-library adoption remain unverified.
• Fifteen earlier local fictional functional cases do not establish production identity, authorization or product acceptance. The later component-specification cases B1–B8 are NOT RUN.
• Eight-session synthetic retrieval observations showed approximately 18.6% more case-source bytes and no avoided case-document reads. No token/cost saving or productivity gain is claimed. The delivery-efficiency comparison remains DEFERRED.
• AI-assisted onboarding is not a clean-machine or cross-provider acceptance result. Earlier live core evidence remains scoped to its original environment/revision.

Validation and immutable source

• Release source: b2980f6876c2a8f9c9ec7ca96e455bc5d9087f15. Git tree: b30eb371da9e606fa42ff13bd466496a5add997a.
• The actual main merge of PR #19 has exactly the tested tree from 741bcc79144e70043fc336e8190572a9a4ef4f35; post-merge commit/parent/tree identity was read back. Identical-tree tests were reused rather than repeated.
• Six existing workflow commands passed locally on Windows/Python 3.12.10 in an isolated standard-library environment: 761 package checks, synchronization regressions, 17 methodology/final-gate checks, the 35-topic audit without errors, 23 catalog/coverage checks and seven onboarding tests.
• Full-text privacy/secret/operational-locator checks, changed-content/metadata/rights review and protected-file comparisons passed. All seven changed remote documents match tested bytes, SHA-256, Git blob IDs and the immutable tree. Code/runtime/inventory/schema/state/bootstrap/pin/license/catalog bytes are unchanged from the preceding approved finalization.
• Earlier publication-guard negative cases and live/fictional observations remain attributed to their original revisions/scopes; no new efficacy result is claimed. PR #19 records actual candidate QA and readback.

Hosted QA disposition

Hosted QA did not execute because Static QA remains manually suspended. Local Windows PASS is not hosted Linux PASS.

The framework maintainer explicitly accepted ADR-0031 on 2026-10-17, limited to the v3.7.0 publication path and all its mandatory evidence conditions. Acceptance was persistently recorded through approved PR #19. This is not a standing waiver, a new runner-zero observation, or a claim that manual suspension satisfies ADR-0022 criteria. Other proposed ADRs and optional-pilot classifications remain unchanged.

Publication is a separate accountable action for this exact source and note text. The exception grants no future-release, hosted-reactivation, private-runtime, Figma or pilot-promotion authority. Restore real hosted evidence when authorized and available; preserve prior limits and failures.

Licensing

Licensing remains PolyForm Noncommercial License 1.0.0 for framework/software-like material and CC BY-NC 4.0 for documentation; commercial use requires a separate written license. This project is source-available for noncommercial use, not OSI Open Source. External concepts were adapted with attribution and stated reuse limits; no complete third-party system was imported.