Skip to content

RELEASES.md

Latest commit

 

History

History
783 lines (518 loc) · 56.7 KB

RELEASES.md

File metadata and controls

783 lines (518 loc) · 56.7 KB
Raw

Automotive Skills Suite — Releases

Weekly snapshots of github.com/jherrodthomas/automotive-skills-suite. Tags are lightweight; the human clicks Publish on GitHub after reviewing this file.

v2026.09.W39 — 2026-09-26

ISO week 39 (2026-09-21 → 2026-09-25). Weekly snapshot, 7 commits.

Highlights

  • cs-goals-builder's hard-coded Cybersecurity_Property = "Confidentiality" (flagged as Known Issue in [v2026.09.W38], opened as #62 Monday) is fixed: the property is now derived per row from the threat's STRIDE category via a new _build_stride_index(), with unmatched rows reporting Not Determined on stderr instead of silently defaulting. Verified across six scenarios including the negative case a hardcode can't fake — an Information-Disclosure-stripped input producing zero Confidentiality rows (#62, 4604f67).
  • The suite-wide _find_header_row gate (#63) was measured down before it was fixed. The issue's own proposed blanket >= 2 threshold would have regressed two already-correct headers (control-plan, ppap), so the fix ships as "prefer ≥3, fall back to 2" instead, proven zero-diff across 2,616 checklist rows built from all 24 paired outputs before/after. Neither of the two symptoms the plan named turned out to move a verdict, so the issue closed HIGH → MED — but the same sweep surfaced two skills that are dead on arrival (safety-case-builder, bus-load-analysis-checklist-reviewer), left unfixed and tracked in Known issues.
  • A new column-level chain-contract audit (#64) caught what the tab-level one couldn't see, and immediately found 13 more instances. scripts/column_contract.py extends the existing audit from tab-existence to column-name and data-start-row assertions, built by first watching it fail on cs-concept-builder's known-bad 02_CS_Goals_Echo edge (#65, 3 of 15 CS Goals silently dropped, every CAL echoing goal text) before fixing it. Run for real against the whole suite, it reports the entire TSC→software branch — sw-arch-builder, sw-sr-builder, sw-fmea-builder — has been running on misread or empty input for an unknown period while the tab-level audit reported every one of those edges MATCH.

Changes this week

plan

  • ea692a7 auto(plan): W39 opens #62-#65, header-row HIGH resized to 24 latent 2 live

polish

  • 4604f67 auto(polish): #62 cs-goals property derived from STRIDE, no longer constant Confidentiality
  • 7796ade auto(polish): journal and STATUS for the #62 STRIDE property fix
  • 50ba884 auto(polish): #63 header gate now prefers 3 cells, falls back to 2; zero-diff proven
  • 5dcce3d auto(polish): #64 column-level chain audit, reproduces #65 and finds three more
  • 6d0118f auto(polish): #65 cs-concept reads CS Goals by header name, three dropped goals restored

docs

  • dcc45fb auto(docs): W39 changelog rolled, 18 checklist-reviewer example stubs, STATUS regen

release (this snapshot commit)

  • STATUS.md regenerated (76/76 paired, 5 fresh / 71 stale / 0 orphan)
  • RELEASES.md appended with this W39 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.09.W39] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76, 2 via docs/PAIRING_ALIASES.md)
  • Freshness: 5 🟢 · 71 🟡 · 0 🔴
  • Domain spread: safety 15, quality 10, comms 8, cyber 6, v&v 5, program-mgmt 5, diagnostics 5, autosar 5, sysml 4, aspice 4, sotif 3, mbse 3, calibration 3
  • Archives touched this snapshot window: 25 skill files (24 checklist-reviewer archives from #63's header-gate batch, plus cs-goals-builder from #62 and cs-concept-builder from #65 — several archives overlap both changes)

Open issues at snapshot

Unknown this run. api.github.com returned 403 on every call this session — read, authenticated or not, against this repo or any other — with the same "not in this session's authorized repository set; use add_repo" body the sibling robotics-skills-suite task hit on its 2026-09-25 DOCS run. This is a session-level proxy restriction, not a repo-state fact, so no issue count is reported rather than guessing. See docs/AUTONOMOUS_LOG.md for today's entry and the human note below.

Known issues carried into this tag

Six items are listed under Known issues in the [v2026.09.W39] CHANGELOG.md section and are not repeated here in full. The two a human should weigh first: three column-level BREAKs from #64 (sw-arch-builder, sw-sr-builder, sw-fmea-builder) are dynamically confirmed HIGH-severity defects not yet filed as issues; and safety-case-builder / bus-load-analysis-checklist-reviewer are both dead on arrival with mechanical, already-written fixes sitting in docs/skill-polish-log/.

For the human

  • GitHub API access is blocked for this scheduled task's cloud-workspace session, and so is git push from that same workspace — this run's commits were made and pushed from the connected desktop instead (see journal). This matches the robotics-skills-suite task's 2026-09-25 report and yesterday's automotive DOCS run exactly: read access over git-HTTPS works from the cloud workspace, but api.github.com and cloud-workspace git push both return 403 from an Anthropic-side proxy that scopes repositories per session. This is the second consecutive day this repo needed the device-side workaround; it will keep needing it until the session's authorized-repository set includes jherrodthomas/automotive-skills-suite for both the API and the push path.
  • Filing needed once the API is reachable: the three #64 column-level BREAKs above, plus the four items still sitting in docs/triage/2026-09-20.md section 4 that W39's plan carried rather than filed.
  • Two tags may still be pushed and unpublished from before this run: v2026.09.W36 and v2026.09.W38 (noted in the W39 plan). Review RELEASES.md and click Publish on whichever tags reached GitHub.

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.09.W38...v2026.09.W39

v2026.09.W38 — 2026-09-19

ISO week 38 (2026-09-06 → 2026-09-18). Accumulating since v2026.09.W36 (2026-09-05) — two weeks in one tag, for the second release running. No v2026.09.W37 tag exists and none is backfilled: the W37 Saturday RELEASE was one of five scheduled runs that did not fire across the fortnight. 7 commits.

Highlights

  • Four builder/reviewer pairs repaired, and every defect was the same shape: a documented input path that accepts analyst data and silently discards it. traceability-matrix (input never reached 11 advertised tabs, reviewer certified the empty result), hw-safety-reqs-builder (hw_sr_overrides dropped entirely on a four-vs-three-segment HW-TSR ID drift, exit 0), cs-goals-builder (TARA columns resolved positionally, so every CSG shipped with the goal text in the Asset column and a risk integer as its goal text), dia-checklist-reviewer (three false findings from probe assumptions about layouts the builder does not emit). None raised an exception; all four had been shipping for months.
  • The verdict on tooling is harsher than the fix list. chain_contract_audit.py reported 0 BREAK across all 16 chains for the entire time the tara-builder → cs-goals-builder payload was corrupt, because it asserts that a referenced tab exists and never that the columns line up. The week's own fixes are the argument for a column-level assertion — it is the single highest-leverage change available and would retro-catch this class suite-wide.
  • Five scheduled runs did not fire (09-09 Wed → 09-14 Mon). W37 lost its Wed/Thu POLISH, Friday DOCS and Saturday RELEASE; W38 lost its Monday PLAN. Consequences carried into this snapshot: two W37 targets (#60 mbse wiring, #61 sheet-name confirmations) were opened, closed completed by the human on 09-11, and never worked — their defects are still in the archives. All three W38 POLISH days selected by rule (c), least-recently-touched, because there was no plan and no open issues.

Changes this week

triage

  • a92088c auto(triage): 12 issues reviewed, 5 stale commented, no low-confidence labels applied

plan

  • f95f4c9 auto(plan): W37 opens #59-#61, obligation hypothesis disproved, sysml batch deferred — the pre-slotting audit killed a hypothesis three previous plans had carried (the Shall/Must REJECTED gate is dead in 3 archives, not 76), collapsing a would-be batch target into a three-line add

polish

  • 302660e auto(polish): #59 traceability pair repaired, 11 tabs wired, empty workbook no longer passes
  • a8f430d auto(polish): hw-safety-reqs overrides silently dropped on TSR ID drift, now applied and warned
  • a2af2f1 auto(polish): cs-goals-builder read TARA by header name, goal text and asset no longer swapped
  • 9170271 auto(polish): dia reviewer misread joint reviews, scope and combined RACI cells

docs

  • 54e46b3 auto(docs): W37+W38 changelog rolled, four example READMEs, five missed runs reconciled — four example READMEs written against the archives, and the first docs roll since W33 to find no check-count drift

release (this snapshot commit)

  • STATUS.md regenerated (76/76 paired, 7 fresh / 69 stale / 0 orphan — fmeda-builder aged out of the 30-day window overnight)
  • RELEASES.md appended with this W38 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.09.W38] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76, 2 via docs/PAIRING_ALIASES.md)
  • Freshness: 7 🟢 · 69 🟡 · 0 🔴
  • Domain spread: safety 15, quality 10, comms 8, cyber 6, autosar 5, diagnostics 5, program-mgmt 5, v&v 5, aspice 4, sysml 4, calibration 3, mbse 3, sotif 3
  • Archives touched this snapshot window: 4 (traceability-matrix-builder, traceability-matrix-checklist-reviewer, hw-safety-reqs-builder, cs-goals-builder, dia-checklist-reviewer — 5 files across 4 pairs)

Open issues at snapshot

0 open. The human closed all 15 on 2026-09-11 after six consecutive plans cited the queue. Two caveats for the reader: #60 and #61 were closed completed without ever being worked, and the four follow-ups raised in the cs-goals-builder and dia-checklist-reviewer polish logs were never filed because the queue tooling was not exercised this fortnight. An empty issue list here means unfiled, not done.

Known issues carried into this tag

Six items are listed under Known issues in the [v2026.09.W38] CHANGELOG.md section and are not repeated here. The two structural ones worth a human's attention before the next plan: _find_header_row requires ≥3 populated cells, which makes every two-column tab in any builder invisible to its paired reviewer; and no regression harness runs builder output through the paired reviewer's probe, which is the exact shape of all four defects fixed this fortnight.

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.09.W36...v2026.09.W38

v2026.09.W36 — 2026-09-05

ISO week 36 (2026-08-24 → 2026-09-05). Accumulating since v2026.08.W34 (2026-08-22) — two weeks in one tag. The scheduled run died after 2026-08-27 and missed the W35 Friday DOCS roll and the W35 Saturday RELEASE, so no v2026.08.W35 tag exists and none is backfilled here.

Highlights

  • Seven reviewers that crashed on every input now run. The single largest correctness fix in the repo's history, and it was one line repeated seven times. sysml-block-diagram-checklist-reviewer (W35 Thu, #56) passed a flat list where dashboard.py:138 expected {tab: [(check, result)]}; the identical line sat in the other three sysml reviewers and, a week later, in all three mbse reviewers (W36 Thu, #58). Both domains went from zero working reviewers to fully working — sysml 0→4, mbse 0→3. Neither defect was findable by reading the archives; both surfaced only because the polish pass executed the reviewer against its own builder's output. That is now the strongest argument on the board for the proposed fifth invariant lint: every reviewer's generate_checklist.py must run to completion on its builder's {} output.
  • sysml-block-diagram-builder is the reference implementation for JSON input (W36 Wed, #57). It takes --input <model.json>, the five hard-coded placeholder blocks are gone from the generator, the schema is documented in SKILL.md, and a brake-by-wire sample_input.json ships with it. The paired reviewer's BDD-12 validates both connector ends instead of one. The W37 sysml batch (activity, requirement, state-machine) reuses these helpers. Behaviour change — see Known issues.
  • autosar closed out. #54 stopped autosar-bsw-config-builder silently dropping memory_layout and bus_interfaces, which made the paired reviewer's C005/C006 reachable for the first time; #55 restored the canonical 5,782-byte recalc.py and corrected the mis-generated ## Skills inventory heading across the remaining five archives. Repo-wide the two mechanical invariants are now clean: recalc.py 152/152 at one hash, bad heading 0/152.
  • Two never-entered domains entered, and both were documented rather than rewritten. traceability-matrix (v&v, #52) is the worst pair found to date — the builder ignores its input and emits five heading cells, and the reviewer certifies that empty workbook at 24/25 compliant because 22 of its 25 checks are a hard-coded LC fall-through. mbse-system-context (#58) reads its input and fills five tabs but leaves seven as single-cell placeholders. Both filed HIGH in the polish logs; neither repaired, because both repairs are authoring, not fixing. Every one of the 13 domains has now had at least one polish pass.
  • Check-count drift is now the repo's largest open honesty problem. Four more reviewers measured this fortnight advertise 28 / 25+ / 30+ / 28+ checks and ship 15 / 6 / 8 / 6. Added to the W33/W34 findings that is roughly 100 advertised checks that do not exist across at least eight reviewers. Standing item 9 — author the missing checks, or downgrade the SKILL.md claims — is no longer a question about one file and now blocks honest documentation on eight.
  • Inventory unchanged at 76 builder + 76 reviewer pairs, 100% paired. No skill added or removed this fortnight.

Changes this snapshot

feat

  • 72eefda auto(polish): #57 sysml-block-diagram gets JSON input, placeholders out, BDD-12 symmetric

fix

  • 84ea8b4 auto(polish): fix #54 bsw-config input drop, C005/C006 live, all mandatory FC
  • 656bdfe auto(polish): fix #55 autosar batch, recalc 152/152, bad heading 0/152
  • 6e8cce6 auto(polish): fix #56 sysml probe crash, 4 reviewers repaired, domain gaps logged
  • da1082a auto(polish): #58 mbse first pass, three crashing mbse reviewers fixed, stakeholder needs row resolved

polish

  • 5660048 auto(polish): #52 traceability pair non-functional, reviewer passes empty workbook

docs

  • 74d017c auto(triage): twelve issues audited, stale rule misfires on two finished issues
  • 6ed7fe5 auto(plan): W35 takes two autosar defects and opens sysml, defers v&v #52
  • 30044b2 auto(plan): W36 slots deferred #52 first, opens sysml input and mbse first pass
  • 357da3d auto(monthly): KPI report for August 2026
  • b5fa14f auto(docs): W35+W36 changelog rolled, four mbse/sysml reviewer stubs, drift logged

release (this snapshot commit)

  • STATUS.md regenerated via scripts/regen_status.py (76/76 paired, 9 fresh / 67 stale / 0 orphan)
  • RELEASES.md appended with this section
  • CHANGELOG [Unreleased] rolled into ## [v2026.09.W36]
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76
  • Reviewers: 76
  • Paired: 76/76 (100.0%, incl. 2 alias pairings per docs/PAIRING_ALIASES.md)
  • Freshness: 🟢 9 touched ≤30d · 🟡 67 stale · 🔴 0 orphan
  • Domain spread: safety=15, quality=10, comms=8, cyber=6, autosar=5, diagnostics=5, program-mgmt=5, v&v=5, aspice=4, sysml=4, calibration=3, mbse=3, sotif=3

Freshness fell 10 🟢 → 9 🟢 overnight with no skill regressing, and the reason is a measurement artifact worth fixing: regen_status.py keys Last Touched on the builder file only. Seven reviewer archives were repaired this fortnight and not one of them moved a row. A max(builder, reviewer) change is one line; it is deferred because the STATUS table is a contract the Monday PLAN and Saturday RELEASE runs both read, and changing what a column means is not a Saturday job.

Open issues at snapshot

12 open: #43, #44, #45, #46, #47, #48, #49, #50, #51, #52, #57, #58. All twelve have met their definition of done. No issue has been closed since #53 on 2026-08-19; the backlog has been purely decorative for three weeks and is actively corrupting the Monday PLAN priority rule, which ranks candidates by "skills referenced by open issues". A human close pass remains the single highest-leverage ten minutes available on this repo, and it is now the second consecutive tag saying so.

Known issues shipped with this tag

  • Behaviour change: generate_sysml_block.py with no --input emits an empty template with a stderr warning, not the old five placeholder blocks. Anything scripting the bare command should point at examples/sysml-block-diagram-builder/sample_input.json
  • Check-count drift, measured from check_definitions.py on 2026-09-04: sysml-block-diagram-checklist-reviewer advertises 28 checks / 7 tabs, ships 15 / 3; mbse-system-context-checklist-reviewer advertises "25+", ships 6; mbse-model-architecture-checklist-reviewer advertises "30+", ships 8; mbse-requirements-allocation-checklist-reviewer advertises "28+", ships 6. Same class as fmeda, autosar-bsw-config, cs-architecture and cdd
  • REJECTED may be dead code suite-wide. dashboard.py counts major issues as NO and obligation == "Shall" while the checks use Must / Should, so a {} workbook scoring 4 NO reports CONDITIONAL APPROVAL. Confirmed in the mbse reviewers; grep suggests the vocabulary split is wider. A repo-wide count across all 76 dashboard.py is queued and is one line per archive if confirmed
  • traceability-matrix pair is non-functional as shipped (above). test-case-catalog (probes Test Case Inventory, builder emits Test Cases) and flexray-config (probes Title, builder emits Title_Page) each have a one-name sheet mismatch that may be a one-line fix — unconfirmed
  • The 2026-09-01 chain scan found builder-to-reviewer breaks, which falsifies the premise on which docs/chain-contract-audit.md excludes those pairs. #46 should be re-scoped, not closed; that decision has been awaiting a human since 09-01

Compare: https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.08.W34...v2026.09.W36

v2026.08.W34 — 2026-08-22

ISO week 34 (2026-08-17 → 2026-08-22). Accumulating since v2026.08.W33 (2026-08-15).

Highlights

  • The five-week carry finally landed, and it paid for itself inside 48 hours. #46 (repo-wide chain-contract audit) had been slotted last and displaced every week since W30. Monday descoped it to builder-to-builder sheet-name reads and slotted it first; it landed Tuesday, immediately found one silent BREAK in fmeda-builder's TSC reader, and that break was fixed Wednesday. Repo chain breaks are now zero. The scheduling lesson — a target that loses every week is a target that is slotted wrong, not a target that is too big — is the most reusable thing in this tag.
  • First pass on autosar, the largest never-polished domain. autosar-bsw-config-builder gained a real examples/sample_input.json (previously the skill's own definition-of-done was unsatisfiable — there was no sample to smoke-test from) and a working recalc.py. Two further defects were found and deliberately not fixed in the same pass; they went out as #54 and #55 rather than being absorbed into a refactor.
  • Check-count drift is now a repo-wide class, not four incidents. Writing example stubs against the unpacked archives rather than the SKILL.md prose surfaced two more reviewers misstating their own check totals: fmeda-checklist-reviewer (claims 28, defines 36) and autosar-bsw-config-checklist-reviewer (claims "~30", defines 9 — off by more than 3×). With cs-architecture (42→43) and cdd that makes four confirmed. A scanner comparing every reviewer's advertised count against len(CHECKS) would find the rest in one pass; it is the leading PLAN candidate for W35.
  • W34 landed 3 of 4, not 4 of 4. #52 (traceability-matrix-builder, v&v) lost Thursday's slot to #53. That was correct triage, but v&v remains a never-polished domain and #52 is a carried target, not a completed one. Recorded here verbatim so the tag does not round it up.
  • Inventory unchanged at 76 builder + 76 reviewer pairs, 100% paired.

Changes this week

plan

  • 547ced0 auto(plan): W34 descopes #46 after five carries, opens autosar and v&v targets

polish

  • 5fa7109 auto(polish): #46 chain-contract audit lands, finds silent fmeda-to-TSC break (#53)
  • ec8fc8e auto(polish): fix #53 fmeda TSC reader, repo chain breaks now zero
  • 15d4c8b auto(polish): #51 autosar-bsw-config gains sample input, drops two silent bugs to issues

docs

  • dd687bd auto(docs): W34 changelog roll, two reviewer stubs expose check-count drift, STATUS regen

release (this snapshot commit)

  • STATUS.md regenerated via scripts/regen_status.py (76/76 paired, 10 fresh / 66 stale / 0 orphan)
  • RELEASES.md appended with this section
  • CHANGELOG [Unreleased] rolled into ## [v2026.08.W34]
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76
  • Reviewers: 76
  • Paired: 76/76 (100.0%, incl. 2 alias pairings per docs/PAIRING_ALIASES.md)
  • Freshness: 🟢 10 touched ≤30d · 🟡 66 stale · 🔴 0 orphan
  • Domain spread: safety=15, quality=10, comms=8, cyber=6, autosar=5, diagnostics=5, program-mgmt=5, v&v=5, aspice=4, sysml=4, calibration=3, mbse=3, sotif=3

Freshness moved 16 🟢 → 10 🟢 in a single day with no skill regressing. Six builders (fsc, hara, pfmea, ppap-package, sw-fmea, tsc) were all last touched on 2026-07-22 in one batch and crossed the 30-day line together. This is a window artifact, not a decline.

Open issues at snapshot

12 open: #43, #44, #45, #46, #47, #48, #49, #50, #51, #52, #54, #55. Ten of these have met their definition of done and remain open only because the autonomous run never closes issues. Genuinely open work: #52 (v&v polish, displaced from Thursday, takes W35's first slot), #54 (autosar-bsw-config drops memory_layout and bus_interfaces silently; builder and reviewer must land in one commit) and #55 (mis-generated template batch, 5 files remaining).

#53 was closed by a human during the week — the first close in five weeks of a backlog that only grew. The other ten DoD-met issues are now actively distorting the Monday PLAN priority rule, which ranks by "skills referenced by open issues". A human close pass is the single highest-leverage ten minutes available on this repo.

Known issues shipped with this tag

New this week — both need a .skill zip repack, i.e. POLISH-mode work:

  • fmeda-checklist-reviewer — SKILL.md says "14 + 14 checks" and labels its Confirmation Review tab 14; check_definitions.py registers 18 CR entries, so the true total is 36 (18 CR + 14 FMEDAA + 4 VA), not 28
  • autosar-bsw-config-checklist-reviewer — SKILL.md and the check_definitions.py docstring both advertise "~30 compliance checks"; the file defines 9 (C001–C009). An analyst reading "~30" reasonably assumes coverage this tool does not have, and two of the nine cannot fire against real builder output at all (#54)

Carried from v2026.08.W33, all still open, all requiring a repack: cdd-checklist-reviewer dangling references/ pointers; cs-architecture-checklist-reviewer count 42→43 / 14→15 VA; the two interacting sotif-analysis-checklist-reviewer probe defects; and the 19 over-length sheet names in docs/sheet-name-length-audit.md.

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.08.W33...v2026.08.W34

v2026.08.W33 — 2026-08-15

ISO week 33 (2026-08-10 → 2026-08-15). Accumulating since v2026.08.W32 (2026-08-08).

Highlights

  • The generator that writes STATUS.md is now version-controlled. For weeks each run re-derived the pairing logic inline, so every fix to it evaporated overnight — Thursday's run duly reported 74 paired / 2 orphan, contradicting docs/PAIRING_ALIASES.md. scripts/regen_status.py landed Friday and parses the alias table out of that doc rather than hard-coding it. Back to 76/76 paired, and it stays that way.
  • Three POLISH landings in three POLISH days, second week running: sotif-analysis-builder (sheet names over Excel's 31-char limit — openpyxl warns and writes through, so shipped workbooks were malformed), cs-architecture-builder (chain reader rewritten against the tabs cs-concept-builder actually emits: 0/0/0 parsed before, 2/6/1 after), and cs-concept-builder (re-verified byte-clean).
  • A prior week's finding was retracted, not quietly dropped. The "13 more corrupt archives" note from the NUL-byte work was a scanner false positive — zip members legitimately end in NUL via the EOCD record. An EOCD-aware rescan puts the true count at 0 corrupt archives out of 152.
  • Domain labels corrected. msa-gage-rr-builder and spc-chart-builder are Gage R&R and SPC — IATF 16949 quality tooling, not ISO 26262 safety. Nine builders matched no documented prefix rule and were hand-assigned each run; all nine are now pinned in the generator. This is why the domain spread below differs from W32's: it is a correction, not a migration.
  • Inventory unchanged at 76 builder + 76 reviewer pairs, 100% paired.

Changes this week

plan

  • fbcf48e auto(plan): W33 carries three targets, opens cdd-builder for diagnostics, STATUS regen

polish

  • 386191e auto(polish): fix Excel-invalid SOTIF sheet names, restore cdd-builder sample fixture
  • b9c8aef auto(polish): repair cs-concept to cs-architecture chain break per #43, STATUS regen
  • 08cf363 auto(polish): #44 verified closed-loop, corrected scanner disproves wider archive corruption

docs

  • dc732bf auto(docs): W33 changelog roll, three reviewer stubs, STATUS generator committed

release (this snapshot commit)

  • STATUS.md regenerated via scripts/regen_status.py (76/76 paired, 15 fresh / 61 stale / 0 orphan)
  • RELEASES.md appended with this section
  • CHANGELOG [Unreleased] rolled into ## [v2026.08.W33]
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76
  • Reviewers: 76
  • Paired: 76/76 (100.0%, incl. 2 alias pairings per docs/PAIRING_ALIASES.md)
  • Freshness: 🟢 15 touched ≤30d · 🟡 61 stale · 🔴 0 orphan
  • Domain spread: safety=15, quality=10, comms=8, cyber=6, autosar=5, diagnostics=5, program-mgmt=5, v&v=5, aspice=4, sysml=4, calibration=3, mbse=3, sotif=3

Open issues at snapshot

8 open, all labeled weekly-target: #43, #44, #45, #46, #47, #48, #49, #50. #43, #44, #45, #47 and #48 have met their definition of done and remain open only because the autonomous run never closes issues — they are queued for a human click. Genuinely carrying into W34: #46 (repo-wide chain-contract audit, carried ×5 and never started), #49 (program-mgmt, DoD met W32 but unclosed), #50 (cdd-builder diagnostics polish, opened Monday, partial — sample fixture restored, reviewer-side dangling references still open).

Known issues shipped with this tag

Four defects were found this week and deliberately left unfixed, because each needs a .skill zip repack (POLISH-mode work):

  • cdd-checklist-reviewer — SKILL.md points at references/methodology.md and references/cdd_checks.md; neither exists in the archive
  • cs-architecture-checklist-reviewer — advertises 42 checks / 14 verification checks; check_definitions.py defines 15 VA checks, so the real total is 43
  • sotif-analysis-checklist-reviewer — insufficiency keyword probe never matches (count permanently 0) and the header row is counted as data; must be fixed together
  • docs/sheet-name-length-audit.md — 19 over-length sheet names across 10 skills still unactioned; worst are secure-coding-guidelines-builder (37) and 8d-problem-solving-builder (34 ×2)

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.08.W32...v2026.08.W33

v2026.08.W32 — 2026-08-08

ISO week 32 (2026-08-03 → 2026-08-08). Accumulating since v2026.08.W31 (2026-08-01).

Highlights

  • Three real polish landings in three POLISH days — the first week since W30 to hit that rate: aspice-process-evidence-builder (input keys now actually render), a2l-builder (comma-string members no longer counted as characters), and safety-program-risk-register-builder (docs corrected from 9 claimed tabs to the true 8).
  • Two of the three closed out long-carried issues: #45 is now fully serviced (all three ASPICE thirds done) and #47's definition of done is met. Both are waiting on a human to click Close.
  • First-ever pass on the program-mgmt domain (#49) — the domain-spread gap flagged in the W32 plan narrowed by one. diagnostics, autosar, mbse, sysml and v&v remain untouched since the May import.
  • Inventory unchanged at 76 builder + 76 reviewer pairs, 100% paired.

Changes this week

plan

  • fcc4c0b auto(plan): W32 carries five open targets, no new issues, STATUS regen

polish

  • f00f1b2 auto(polish): aspice-process-evidence tabs 06/07/08 wired to real input data
  • 6fab2ce auto(polish): a2l-builder members normalization and conversion_methods schema doc fix
  • 9b4660e auto(polish): safety-program-risk-register docs made honest, dangling reference removed

docs

  • cd6a29e auto(docs): W32 changelog roll, three example stubs, STATUS regen

release (this snapshot commit)

  • STATUS.md regenerated (76/76 paired, 12 fresh / 64 stale / 0 orphan)
  • RELEASES.md appended with this section
  • CHANGELOG [Unreleased] rolled into ## [v2026.08.W32]
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76
  • Reviewers: 76
  • Paired: 76/76 (100.0%, incl. 2 alias pairings per docs/PAIRING_ALIASES.md)
  • Freshness: 🟢 12 touched ≤30d · 🟡 64 stale · 🔴 0 orphan
  • Domain spread: safety=11, other=9, comms=8, cyber=6, quality=5, autosar=5, diagnostics=5, program-mgmt=5, v&v=5, aspice=4, sysml=4, calibration=3, mbse=3, sotif=3

Open issues at snapshot

7 open, all labeled weekly-target: #43, #45, #46, #47, #48, #49 — plus #44. #44, #45 and #47 have met their definition of done and are only open because the autonomous run never closes issues. Carrying into W33: #43 (cs-concept → cs-architecture reader rewrite, carried x4), #46 (repo-wide chain-contract audit, carried x4), #48 (sotif first pass, carried x2).

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.08.W31...v2026.08.W32

v2026.05.W20 — 2026-05-16

ISO week 20 (2026-05-11 → 2026-05-16). First weekly snapshot of the autonomous run cadence.

Highlights

  • Three builders moved from "shipped" to "reviewed against the polish playbook": hara-builder, cs-concept-builder, aspice-assessment-builder. Each now carries a dated entry in docs/skill-polish-log/ with severity-rated findings.
  • Weekly target file docs/weekly/WEEK-2026-W20.md opened the loop on Monday and the four GitHub issues it spawned are now mostly serviced; one (8d-problem-solving-builder) carries forward to W21.
  • Suite still at 76 builder + 76 reviewer pairs, 100% paired (item-definition-builder ↔ item-def-checklist-reviewer and ppap-package-builder ↔ ppap-checklist-reviewer are alias pairs and now reflected correctly in STATUS).

Changes this week

plan

  • 16df7a5 auto(plan): W20 targets — hara, cs-concept, aspice-assessment, 8d

polish

  • 5b6ad4e auto(polish): regen STATUS.md and log W20 hara-builder polish findings
  • 49fd9a9 auto(polish): regen STATUS.md and log W20 cs-concept-builder polish findings
  • c4fad8a auto(polish): W20 #5 aspice-assessment-builder pass, regen STATUS.md

docs / release (this snapshot commit)

  • STATUS.md refreshed (no flag changes vs. prior run)
  • RELEASES.md created
  • docs/AUTONOMOUS_LOG.md updated with RELEASE-mode entry

Skills inventory

  • Builders: 76
  • Reviewers: 76
  • Paired: 76/76 (100.0%)
  • Domain spread: safety=12, quality=8, comms=8, program-mgmt=6, cyber=6, autosar=5, diagnostics=5, v&v=5, aspice=4, sysml=4, mbse=3, sotif=3, calibration=3, other=4

Open issues at snapshot

  • #6 W20 polish target: 8d-problem-solving-builder.skill (carries to W21)
  • #5 W20 polish target: aspice-assessment-builder.skill (serviced this week, awaiting close-out)
  • #4 W20 polish target: cs-concept-builder.skill (serviced this week, awaiting close-out)
  • #3 W20 polish target: hara-builder.skill (serviced this week, awaiting close-out)
  • #2 goodd (needs human triage — non-conforming title, no domain signal)

Compare

3c69553...v2026.05.W20

Notes for the human

Click Publish on the v2026.05.W20 tag on GitHub once you've skimmed this section. Issue #2 (goodd) is the one item Sunday's TRIAGE pass should probably still escalate to you — autonomous run is intentionally not labeling or commenting it because confidence is well below 80%.

v2026.05.W21 — 2026-05-23

ISO week 21 (2026-05-18 → 2026-05-23). Second weekly snapshot — W21 polish cycle closed; CHANGELOG and the first examples/ doc stubs landed.

Highlights

  • W21 polish ran the full Tue/Wed/Thu cadence over 8d-problem-solving-builder, dbc-builder, and autosar-swc-builder; the autosar-swc pass shipped an actual in-allowlist description fix, the other two produced severity-rated polish-log findings held for human review.
  • CHANGELOG.md introduced and the first examples/<skill>/README.md stubs added — the repo now has a docs spine beyond STATUS.
  • Suite steady at 76 builder + 76 reviewer pairs, 100% paired; no flag changes vs. W20.

Changes this week

plan

  • 3320c23 auto(plan): W21 targets — 8d carryover, dbc, autosar-swc, uds-services, classifier-freeze

polish

  • e8ff3b2 auto(polish): W21 #6 8d-problem-solving-builder pass, regen STATUS.md
  • 76b2fff auto(polish): W21 #7 dbc-builder pass, regen STATUS.md
  • 387bbcd auto(polish): W21 #8 autosar-swc-builder pass — description fix applied, STATUS regen

docs

  • 1a5ca80 auto(docs): add CHANGELOG, W21 example README stubs, regen STATUS

docs / release (this snapshot commit)

  • STATUS.md refreshed (no flag changes vs. prior run)
  • RELEASES.md updated with the W21 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.05.W21] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76)
  • Domain spread: safety 14, quality 8, comms 8, program-mgmt 6, cyber 6, autosar 5, diagnostics 5, v&v 5, aspice 4, sysml 4, calibration 3, mbse 3, sotif 3, other 2

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.05.W20...v2026.05.W21

Open items for the human

  • 10 issues open. Issue #2 ("goodd", empty body) has stayed needs-triage across multiple runs — needs a human decision (likely close as spam/invalid).
  • Issue #11 ("packaging utility for Claude-compatible skill exports") is still unlabeled — Sunday TRIAGE should label it.
  • W21 polish target #9 (uds-services-builder) and tooling target #10 (classifier freeze) were not serviced this week — both carry to W22.

v2026.05.W22 — 2026-05-30

ISO week 22 (2026-05-25 → 2026-05-30). Third weekly snapshot — W22 polish cycle closed, alias map landed in the STATUS regen helper, RELEASE cadence now three weeks running.

Highlights

  • W22 polish ran the Tue/Wed/Thu cadence over uds-services-builder, dfmea-builder, and hara-builder; the uds-services pass shipped an in-allowlist description rewrite (DIDs, RIDs, SecurityAccess, NRC, P2/P2*/S3 timing, three session types, 0x10-0x86 service range, casual phrasings, sibling redirects) and finally cleared the W21-era classifier blind spot. The other two produced severity-rated polish-log entries held for human review.
  • The reviewer-name alias map (item-definition ↔ item-def, ppap-package ↔ ppap) was promoted from inline override into the STATUS regen helper on Fri DOCS, so the 100% paired headline is now data-driven rather than hand-patched per run.
  • Three new examples/<skill>/README.md stubs landed (uds-services, hara, dfmea) bringing the docs spine to seven seeded skills. CHANGELOG groups today's W22 work cleanly into Polish (3) and Docs (1 plan + 1 docs commit).
  • Suite steady at 76 builder + 76 reviewer pairs, 100% paired; no flag changes vs. W21.

Changes this week

plan

  • 973c075 auto(plan): W22 targets — uds, hara, cs-concept, aspice-assessment, dfmea

polish

  • 9850780 auto(polish): W22 #1 uds-services-builder pass, regen STATUS
  • 04482aa auto(polish): W22 #5 dfmea-builder pass, regen STATUS, journal
  • ef78172 auto(polish): W22 #2 hara-builder pass, regen STATUS, journal

docs

  • f9e63f9 auto(docs): W22 changelog roll-up, 3 example stubs, STATUS regen

docs / release (this snapshot commit)

  • STATUS.md refreshed (no flag changes vs. prior run; 76/76 paired)
  • RELEASES.md updated with the W22 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.05.W22] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76)
  • Domain spread: safety 13, other 9, comms 8, cyber 6, autosar 5, diagnostics 5, quality 5, v&v 5, aspice 4, sysml 4, calibration 3, mbse 3, program-mgmt 3, sotif 3

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.05.W21...v2026.05.W22

Open items for the human

  • 10 issues open. The five W22-touched targets (#3 hara, #4 cs-concept, #5 aspice-assessment, #9 uds-services, #12 dfmea) now all carry W22 polish-log evidence — recommend a sweep to close the ones the human considers done.
  • Issue #10 (classifier freeze) is now 7+ runs old; this week's Fri DOCS pass got the alias map into the helper but the helper is still inline-Python in each commit. Mon W23 PLAN should allocate a tooling slot to extract scripts/classify_skill.py.
  • Issue #11 (skill packaging utility) and the W22 carryover targets (cs-concept, aspice-assessment, which only have W20-dated polish-log entries) are the natural W23 candidates if the human wants visible motion next week.

v2026.06.W23 — 2026-06-06

ISO week 23 (2026-06-01 → 2026-06-06). Fourth weekly snapshot — first of June. Tag scheme ruling defaulted to ISO-absolute (W23, continuing the W20/W21/W22 series) after four unanswered flags; per-month spelling can be re-cut from the same SHA if a maintainer prefers.

Highlights

  • W23 polish cycle completed its three passes: cs-concept-builder (W20 findings re-confirmed), tara-builder (1 medium finding — Auto-rating Heuristics internal contradiction), and fmeda-builder (2 medium findings — Classification ladder unreachable branch; SMvDU non-standard acronym — plus a suspected 100× unit-convention discrepancy in the JSON example). None of these were applied — they are polish-log findings carried into the W24 maintainer backlog, not fixes in this snapshot.
  • May 2026 monthly KPI report published (23 commits, 24 distinct skills touched, 3 weekly releases, 100% paired ratio; SOTIF flagged as the only zero-touch domain in May).
  • Example-stub coverage rose from 7.9% (6/76) to 11.8% (9/76) with new stubs for cs-concept, tara, and fmeda.
  • Suite steady at 76 builder + 76 reviewer pairs, 100% paired; canonical domain classifier (Tue 2026-06-02 explicit map) re-applied, spread stable at safety=15 / quality=10.

Changes this week

plan

  • 3af1f6b auto(plan): W23 targets — cs-concept, aspice-assessment, classifier, fmeda, tara

polish

  • 5b4b006 auto(polish): W23 #1 cs-concept-builder pass, regen STATUS, journal
  • 22d6409 auto(polish): W23 #2 tara-builder pass, regen STATUS, journal
  • d6afa26 auto(polish): W23 #3 fmeda-builder pass, regen STATUS, journal

docs

  • f8e940f auto(monthly): KPI report for May 2026
  • c21a84a auto(docs): W23 changelog roll, 3 example stubs, STATUS regen

docs / release (this snapshot commit)

  • STATUS.md refreshed (76/76 paired; canonical spread; no flag changes vs. Fri)
  • RELEASES.md updated with the W23 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.06.W23] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76)
  • Domain spread: safety 15, quality 10, comms 8, cyber 6, autosar 5, diagnostics 5, program-mgmt 5, v&v 5, aspice 4, sysml 4, calibration 3, mbse 3, sotif 3

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.05.W22...v2026.06.W23

Open items for the human

  • 12 issues open. The W20-era carryovers (#4 cs-concept, #5 aspice-assessment) have carried ready-to-apply rewrites in the polish log for 3+ weeks — recommend converting both to maintainer PRs in W24 PLAN rather than re-queuing as polish targets.
  • The fmeda medium findings (Classification ladder unreachable branch; SMvDU acronym needing an ISO 26262-5 Annex B source-check; 100× distribution_pct unit suspicion) are the highest-leverage maintainer items from W23 — see docs/skill-polish-log/fmeda-builder.md.
  • Issue #10 (classifier extraction to scripts/classify_skill.py) is now 8 consecutive runs of hand-maintained inline Python — the explicit map grew again this week (item-definition → safety).
  • Issue count dropped 13 → 12 between Thu and Fri without autonomous action; presumed human/spam-filter close of #17 — Sun TRIAGE will confirm.

v2026.06.W25 — 2026-06-20

ISO week 25 (2026-06-15 → 2026-06-20). This snapshot consolidates two weeks of work: the W24 Saturday release (2026-06-13) did not execute, so everything committed since the last tag (v2026.06.W23, 2026-06-06) ships here — the W24 plan/triage/polish plus the W25 docs roll. Internal commit messages carry "W24" labels (a one-week drift between the human's week labels and ISO week numbers); the tag follows the established ISO-week-of-year convention.

Highlights

  • Three polish passes landed against sotif-analysis-builder, safety-case-builder, and control-plan-builder. Only control-plan-builder carried an embedded .skill fix (JSON example sample-count corrected 18 → 15 to match its 15 controlled characteristics); the other two are polish-log entries with no .skill edits.
  • The W24 plan opened the loop on five targets (classifier extraction, sotif, safety-case, control-plan, comm-matrix); a triage pass typed #12 and confirmed the #17 delta. All issues have since been serviced or closed — the tracker is at 0 open at snapshot.
  • Suite holds at 76 builder + 76 reviewer pairs, 100% paired. No orphans, no new skills this period.

Changes this week

plan

  • 2623e14 auto(plan): W24 targets — classifier extract, sotif, safety-case, control-plan, comm-matrix

polish

  • a6df28d auto(polish): W24 #20 sotif-analysis-builder review pass, regen STATUS, journal
  • 2bd0c19 auto(polish): W24 #21 safety-case-builder review pass, regen STATUS, journal
  • 54e2559 auto(polish): control-plan-builder pass, fix sample count 18->15, regen STATUS, journal

triage

  • d8b19ef auto(triage): label refresh, #12 typed, #17 delta confirmed, STATUS regen

docs

  • 6051a5d auto(docs): backfill W24 changelog, add 2 example stubs, regen STATUS

docs / release (this snapshot commit)

  • STATUS.md refreshed (76/76 paired; no flag changes vs. Fri — only the generated-on date advanced)
  • RELEASES.md updated with this W25 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.06.W25] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76)
  • Domain spread: safety 17, quality 10, comms 8, cyber 6, autosar 5, diagnostics 5, v&v 5, aspice 4, sysml 4, calibration 3, mbse 3, program-mgmt 3, sotif 3

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.06.W23...v2026.06.W25

Open items for the human

  • Missed W24 release — the 2026-06-13 Saturday run did not fire; W24 work is captured here rather than under its own tag. If a continuous tag history matters, a retroactive v2026.06.W24 could be cut at 6051a5d, but that is a manual call left to the maintainer.
  • Week-label drift — commit messages and weekly docs use "W24" for work done in ISO week 25. Worth deciding whether to align the human labels to ISO weeks or document the offset, so future tags don't look like they skip a number (W23 → W25).
  • Issue #10 (classifier extraction to scripts/classify_skill.py) remains open conceptually — STATUS regen still relies on parsing the prior STATUS for the domain map rather than a committed classifier. No feat: work was scheduled this period.

v2026.06.W26 — 2026-06-27

ISO week 26 (2026-06-22 → 2026-06-27). Saturday RELEASE run; W25 release fired last week, so this is a normal single-week snapshot with no backfill drift.

Highlights

  • A full W26 polish rotation landed: 5-why-builder, communication-matrix-builder, and aspice-assessment-builder each took a review pass. Only aspice-assessment-builder carried an embedded .skill fix (capability-level rating rule corrected to ISO/IEC 33020); the other two are description/log touches.
  • Week cadence is back in sync — commit labels, weekly docs, and the ISO week all read "W26" this period, closing the W23→W25 numbering drift called out in the W25 release notes.
  • Suite holds at 76 builder + 76 reviewer pairs, 100% paired. No orphans and no new skills this period; the tracker is at 0 open at snapshot.

Changes this week

plan

  • e2162f8 auto(plan): W26 targets — 5-why, aspice-assessment, cs-concept, dia, comm-matrix

polish

  • 107d490 auto(polish): 5-why-builder pass — add example stub, polish log, regen STATUS
  • 9389a13 auto(polish): broaden communication-matrix-builder trigger, log, regen STATUS
  • fd66c47 auto(polish): fix aspice-assessment-builder CL rating rule to ISO/IEC 33020

docs

  • 57dd9ba auto(docs): W26 changelog roll, 2 example stubs, STATUS regen

docs / release (this snapshot commit)

  • STATUS.md refreshed (76/76 paired; no flag changes vs. Fri — only the generated-on date advanced)
  • RELEASES.md updated with this W26 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.06.W26] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76)
  • Domain spread: safety 17, quality 10, comms 8, cyber 6, autosar 5, diagnostics 5, v&v 5, aspice 4, sysml 4, calibration 3, program-mgmt 3, mbse 3, sotif 3

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.06.W25...v2026.06.W26

v2026.07.W27 — 2026-07-04

ISO week 27 (2026-06-29 → 2026-07-04). First release of July; normal single-week snapshot, cadence in sync (plan, polish ×3, monthly KPI, docs roll all landed on schedule).

Highlights

  • Full W27 polish rotation completed with three review passes: apqp-plan-builder (smoke-tested, clean), item-definition-builder (med finding: reviewer naming mismatch item-def-checklist-reviewer vs builder base — Option B alias-map recommended, needs human call), and aspice-gap-analysis-builder (med finding: assessment-xlsx argument accepted but ignored; two fix options logged, deferred as cross-file). No .skill edits this week — all findings logged for human decision.
  • June 2026 monthly KPI report published mid-week (docs/), alongside the W27 weekly plan targeting odx, autosar-bsw-config, mbse-context, sysml-state, and traceability builders.
  • Suite holds at 76 builder + 76 reviewer pairs, 100% paired, 0 orphans, tracker at 0 open at snapshot.

Changes this week

plan

  • 8847b29 auto(plan): W27 targets — odx, autosar-bsw-config, mbse-context, sysml-state, traceability

polish

  • 2df3374 auto(polish): apqp-plan-builder smoke-tested, polish log added, STATUS regen
  • 01f38c5 auto(polish): item-definition-builder reviewed, naming-mismatch logged, STATUS regen
  • 260c03f auto(polish): aspice-gap-analysis-builder smoke-tested, ignored-arg finding logged, STATUS regen

docs

  • 35be610 auto(monthly): KPI report for June 2026
  • 19c4faa auto(docs): W27 changelog roll, 3 example stubs, STATUS regen

docs / release (this snapshot commit)

  • STATUS.md refreshed (76/76 paired, 3 fresh / 73 stale; generated-on date advanced to 2026-07-04)
  • RELEASES.md updated with this W27 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.07.W27] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76)
  • Domain spread: safety 17, quality 10, comms 8, cyber 6, autosar 5, diagnostics 5, v&v 5, aspice 4, sysml 4, calibration 3, program-mgmt 3, mbse 3, sotif 3

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.06.W26...v2026.07.W27

v2026.07.W29 — 2026-07-18

ISO week 29 (2026-07-13 → 2026-07-18). Two-week snapshot: the W28 Sat RELEASE run (2026-07-11) never fired — no v2026.07.W28 tag exists — so this release also carries the back-filled W28 work (see CHANGELOG note). Scheduler gap 2026-07-09 → 2026-07-14 flagged for human review.

Highlights

  • Three polish passes this week: aspice-improvement-plan-builder (ignored-arg chain-break logged — required gap_analysis.xlsx CLI arg never read), aspice-process-evidence-builder (clean input-consumption data point; scaffold-only tabs 06–08 logged), and cs-architecture-builder (smoke-tested clean, but a high-severity silent chain-break found upstream: cs-concept-builder's output tabs/headers don't match cs-architecture-builder's reader — a real concept workbook reads back zero CSRs). All findings logged; no .skill edits.
  • Friday DOCS roll back-filled the missed W28 changelog entries (ppap/item-def polish, W28 plan, docs/PAIRING_ALIASES.md canonical alias registry) and added 4 example README stubs.
  • Suite holds at 76 builder + 76 reviewer pairs, 100% paired (2 via alias registry), tracker at 0 open at snapshot.

Changes this week

polish

  • 26ee19a auto(polish): aspice-improvement-plan-builder ignored-arg chain-break logged, STATUS regen
  • 696e912 auto(polish): aspice-process-evidence-builder scaffold-tab findings logged, args clean
  • 8304e58 auto(polish): cs-architecture-builder verified; cs-concept chain-break found and logged

docs

  • 78cfb27 auto(docs): W29 changelog roll with W28 back-fill, 4 example stubs, STATUS regen

docs / release (this snapshot commit)

  • STATUS.md regenerated (76/76 paired, 4 fresh / 72 stale; ISO 26262 core builders dia/fmeda/fsc/hsi/item-definition reclassified other → safety per domain rules)
  • RELEASES.md updated with this W29 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.07.W29] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76, 2 via docs/PAIRING_ALIASES.md)
  • Domain spread: safety 15, comms 8, cyber 6, autosar 5, diagnostics 5, other 5, program-mgmt 5, quality 5, v&v 5, aspice 4, sysml 4, calibration 3, mbse 3, sotif 3

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.07.W27...v2026.07.W29

v2026.07.W30 — 2026-07-25

ISO week 30 (2026-07-20 → 2026-07-25). Weekly snapshot, 5 commits.

Highlights

  • NUL-corruption campaign: the W30 plan flagged cs-concept-builder's generator carrying 3,361 trailing NUL bytes (#44, fixed Tue 691f614); the same pass surfaced 13 more corrupt archives repo-wide, all cleaned in Wednesday's batch (8a855cf) — 9 generators compile again and prior xlsx false-positive findings were corrected.
  • a2l-builder repaired (#47, Thu 661c4bb): real TypeError crash in axis-points handling fixed (type-tolerant len() guard), frontmatter trigger strengthened, 13-tab output smoke-verified.
  • Friday DOCS roll added 11 example README stubs — including the repo's first reviewer-side stubs (flagged for human confirmation). W30 targets #44 and #47 met; #43 (cs chain repair), #45 (ASPICE wiring), #46 (chain-contract audit) carry into W31.

Changes this week

plan

  • a0c3d91 auto(plan): W30 targets — cs chain repair, NUL-strip, aspice bundle, audit, a2l

polish

  • 691f614 auto(polish): cs-concept NUL-strip fixed per #44; 13 more corrupt archives found
  • 8a855cf auto(polish): batch NUL-strip 13 archives, 9 generators compile again, xlsx false-positives corrected
  • 661c4bb auto(polish): a2l-builder axis points crash fixed, trigger strengthened, 13 tabs verified

docs

  • f92275c auto(docs): W30 changelog roll, 11 example stubs, STATUS regen

docs / release (this snapshot commit)

  • STATUS.md regenerated (76/76 paired, 8 fresh / 68 stale; aspice-assessment-builder aged past the 30-day line)
  • RELEASES.md updated with this W30 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.07.W30] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76, 2 via docs/PAIRING_ALIASES.md)
  • Domain spread: safety 15, comms 8, cyber 6, other 5, quality 5, autosar 5, diagnostics 5, program-mgmt 5, v&v 5, aspice 4, sysml 4, calibration 3, mbse 3, sotif 3

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.07.W29...v2026.07.W30

v2026.08.W31 — 2026-08-01

ISO week 31 (2026-07-27 → 2026-08-01). Weekly snapshot, 4 commits. First tag to cross a month boundary — the week opened in July and closes in August, so the tag carries the 2026.08 month prefix of its release date while retaining ISO week 31.

Highlights

  • ASPICE ignored-arg class cleared in two of three builders (#45): aspice-gap-analysis-builder (Tue 67c9072) and aspice-improvement-plan-builder (Thu 299cdad) each dropped a required-but-never-read CLI argument, with docs corrected to describe what the generator actually consumes. Both fixes are deliberately identical in shape so the ASPICE CLIs stay consistent with each other.
  • #45 cannot close yet. The third leg, aspice-process-evidence-builder, has no ignored-arg defect — its input is genuinely consumed. Its only remaining finding is the scaffold-tab question (are Roadmap / Resources / Risks / KPIs / Communication / Pilot tabs input-driven or intentional fill-in templates?), which is a cross-builder schema decision first raised 2026-07-15 and still awaiting a human answer.
  • Thin week, honestly reported. Wednesday 2026-07-29 produced no commit at all, breaking the always-commit invariant; the Friday DOCS run flagged it and the cause is still unknown. Two of the five W31 targets (#48 sotif, #49 program-mgmt — both first passes on never-polished domains) were never started, and #43 (the largest carried item) remains untouched for a second week.

Changes this week

plan

  • 57238bf auto(plan): W31 targets — carry cs chain, aspice, audit; add sotif, program-mgmt

polish

  • 67c9072 auto(polish): aspice-gap-analysis ignored assessment arg removed, docs made honest
  • 299cdad auto(polish): aspice-improvement-plan ignored gap-analysis arg removed, docs made honest

docs

  • e749617 auto(docs): W31 changelog roll, 2 reviewer example stubs, STATUS regen

release (this snapshot commit)

  • STATUS.md regenerated (76/76 paired, 10 fresh / 66 stale / 0 orphan — unchanged from Friday apart from the header)
  • RELEASES.md appended with this W31 section
  • CHANGELOG.md [Unreleased] rolled into a dated [v2026.08.W31] section
  • docs/AUTONOMOUS_LOG.md updated with the RELEASE-mode entry

Skills inventory

  • Builders: 76 · Reviewers: 76 · Paired ratio: 100.0% (76/76, 2 via docs/PAIRING_ALIASES.md)
  • Freshness: 10 🟢 · 66 🟡 · 0 🔴
  • Domain spread: safety 11, other 9, comms 8, cyber 6, autosar 5, diagnostics 5, program-mgmt 5, quality 5, v&v 5, aspice 4, sysml 4, calibration 3, mbse 3, sotif 3

Open issues at snapshot

7 open — #43, #44, #45, #46, #47, #48, #49. #44 and #47 have met their definition of done and are waiting on a human to close them (the run never closes issues autonomously).

Compare

https://github.com/jherrodthomas/automotive-skills-suite/compare/v2026.07.W30...v2026.08.W31