Housekeeping: skill guide published, jev plan on main, openspec tree un-ignored and archived, one-clock fix by NetDevAutomate · Pull Request #35 · NetDevAutomate/StudyLoop

NetDevAutomate · 2026-09-23T07:20:59Z

Summary

The housekeeping pass that follows #32: every loose end the tracker did not hold, each with a countable finish, as one pure fast-forward of main. Twelve commits, one per concern.

#	Concern	Finish	Commits
1	The `studyloop-study-notes` skill work that sat uncommitted on the `main` checkout	On the branch unchanged (cherry-picked from `feat/study-notes-skill`, file-by-file identical; only the CHANGELOG merges under one `### Added`), and its guide published on the docs site as Lesson Study Notes — one exclude line, one nav entry, the SKILL.md link now a GitHub URL so `--strict` accepts it	`a82fa2fd` `2fcdb510` `42d28de3`
2	The two graphify files (`.graphify-labels.json`, `.graphifyignore`) — deletions the owner confirmed deliberate	Untracked; `graphify_refine.py`'s two docstrings no longer claim the labels file is tracked	`2fcdb510`
3	The unpushed `feat/jev-judge` planning branch (Stage 0 receipt + council-validated plan)	Both commits cherry-picked, content verified file-by-file against the originals (only `.secrets.baseline`'s `generated_at` differs — regenerated by a whole-repo `detect-secrets scan`, results file count 72 → 72). Plan §7's five owner decisions filed as #34 (decision 5, merge order, is resolved by this very cherry-pick). Worktree removed by a script that gated the removal on `status --porcelain` empty and every ignored entry being a known cache	`728600c1` `ce78bf98`
4	`openspec/` ignored in `.gitignore` while 78 tracked files live under it	Rule removed. Found because `git add -A openspec/changes` refused the new archive; `git status --ignored` showed nothing else hidden under the path	`7ed7ae3d`
5	The `herdr-ghostty-multiplexer-transport` change: `deferred:` since 2026-09-05, never archived, five open boxes	Archived as `2026-09-23-…` with `--skip-specs`. Each open box carries a `closed 2026-09-23 — NOT built, deferred with the change` disposition and a reopen condition. Deltas not merged, verified against the tree: all eight `MODIFIED` headers exist in no main spec; the `web-ui` delta preserves the wterm selector `e9cb5656` removed and a bootstrap `e4b17d46` replaced; the `session-transports` delta routes a ttyd fallback the PTY start path calls retired. Main specs byte-identical (sha256 of all 23). `openspec/changes` now holds only `archive/`	`fb999c53`
6	The frozen-clock gap recorded in receipt `now-rubric-2026-09-16` row 3c (e)	RED → GREEN. `spaced_repetition_due()` gains keyword-only `now`; `_due_progress_candidates` takes the engine's instant; `build_now_plan` passes its one read. The RED patches `history.progress.datetime` with a decoy (2031) so it fails on any date if the count comes from anywhere but the engine. Deliberate test change: six collector stubs accept `**_`. One RED expectation corrected at GREEN (score = 138 + `PLAN_RELATED_BIAS`, asserted from the constant). Receipt row 3c records the fix beside the gap	`3b6e8e00` `e126973d`
7	The July archive `2026-07-12-complete-e2e-harness-and-desktop-mcp`: four open boxes, the one red in `openspec validate --archived --all` since the release guard was added	All four reconciled against the tree — each was built the same day its "not done" note was written (fake agent `tests/_fake_agent.py`; journey phases across two e2e files; session-end/review rows asserted; MCP explorer tools in `9f09cf08`). `--archived --all` 8/8 for the first time. The Justfile comment and `validate_new_archives` docstring that cited the July case now state the design reason and record it closed	`6d3c43d6` `d277286f`
8	CHANGELOG	`[Unreleased]` Added/Changed/Fixed for the above	`c1a28de1`

Verification (all on this head)

ruff check clean · ruff format --check 1044 files · pyright 0 errors · bandit passed
mkdocs build --strict exit 0 (the study-notes page present in the built site)
openspec validate --specs --all 23/23 · --archived --all 8/8 · scripts/check-release-consistency.py --skip-wheel --release --pre-tag passes (validates the herdr archive as a new archive)
JS node --test 164/164
Engine/decision/history/recap/web-now/web-history 153 passed (RED flipped); docs contract 31; docs drift + counts + second-brain 318; release-consistency script tests 16
Golden now_plan_no_active.json ec451ce8 byte-identical (the no-plan payload is untouched)
Full pytest -rfE: see the comment below (run against the committed environmental id set from the item-4 control receipt)

Owner steps after merge

git merge --ff-only origin/chore/housekeeping-0.5.x && git push origin main (pure fast-forward of 12 commits).
Delete two local branches that this PR supersedes and whose commits it carries with different SHAs: git branch -D feat/study-notes-skill feat/jev-judge — -D because the cherry-picks changed the SHAs, so -d's merged-check refuses; content proven present above.
Say the word and I delete the three remote branches (feat/body-double-first-move, feat/study-notes-skill, chore/housekeeping-0.5.x) in one ruleset window, proving the deletion command is permitted against the live ruleset first.

Not in this PR, on purpose

The 0.5.1 version bump and release notes (the release change's job).
Plan integration follow-on (D-E): age-aware nudge and a learner-issued retire/snooze for an overdue item #26, Promote Pi and OpenCode from preview to core harness tier — with live release evidence, then docs #21, Plan integration follow-on: the low-energy day is a ladder — a step-up offered after a completed floor task, never led with #31, Read-along mode: the co-study companion reads a lesson aloud in short conversational chunks and pauses for questions #33 — real builds, each with a definition of done in its body.
Learning-tier plan (jev-judge): four owner decisions before S1-0 starts #34 — the four owner decisions the jev plan needs before S1-0.
Design decision 10 (high-energy warm-up) stands as built unless the owner reverses it.

One Markdown note per lesson plus a linked section overview, with the seven-part structure of the comprehensive study notes kept intact and explicit Obsidian-only, xTiles-only, dual-destination and portable workflows. The package is skills/studyloop-study-notes/ (SKILL.md, two templates, four references) and has no runtime dependency on StudyLoop; it is installed separately through the skills CLI. This work was written on the main checkout while PR #32 was open and is moved here unchanged so main can fast-forward to the PR head: main and the branch both rewrite the CHANGELOG [Unreleased] section, and the uncommitted copy would have refused the merge. The guide is source-only for now, as mkdocs.yml's documentation contract requires until a page is rooted and added to nav.

Both files were already listed in .gitignore; deleting the tracked copies completes that decision (owner, 2026-09-22: deliberate). They are local graphify tuning — curated community labels and an ignore list — that belong to one machine's graph build, not the repository. scripts/graphify_refine.py already tolerates the labels file being absent (load_curated_labels returns {} when it does not exist); only its docstrings still claimed the file was tracked in git, so they now say local and optional. Nothing else in the tree references either file.

The guide moved onto this branch source-only: mkdocs.yml excludes every page not whitelisted, so it built into nothing until it was both excluded and rooted in nav. This is the skill's finish as the housekeeping plan named it: one exclude line, one nav entry beside Obsidian Export, and the repo-relative link to SKILL.md becomes the GitHub URL because a published page may not link outside the docs tree under --strict. The page itself is unchanged and still says the skill is installed with the skills CLI, not by studyloop install agents. mkdocs build --strict clean, the page present in the built site; test_docs_drift + test_docs_no_hardcoded_test_counts + test_second_brain_docs 318 passed, 3 skipped.

… dimension blind Evaluate TypeSafe's Jev (jev-1.13.0, a System One judgement model) as an optional judgement provider for StudyLoop's unfilled judgement slots. Stage 0 is an access-and-shape check only: one synthetic teach-back with a planted factual error, five Score questions whose criteria are the verbatim 1-4 level descriptors from agents/shared/teach-back-protocol.md, two Noul controls, three identical calls. Why this is recorded rather than thrown away: the probe surfaced a hard constraint early. Own words / structure / depth landed where a human would put them (confidence >= 0.87), but accuracy put 82% of its mass on the two "accurate" levels despite the planted error, and the error Noul sat at 0.43-0.48. Jev is a common-sense judge, not a Python-semantics checker. Its confidence signal did flag accuracy as the least certain dimension (0.52), so a confidence gate would have routed it to "ask the learner". Design consequence carried into Stage 2: the harness LLM supplies accuracy, Jev scores the four pedagogical dimensions, argmax-level + mass gate instead of rounding the float. Identical calls moved <= 0.06 on a 0-3 scale: consistent, not deterministic, so later tests assert bands/levels and CI replays fixtures, never calls live. Out of scope by design: learning/decision.py and the completion review (counting + dates), which the vendor's jaggedness page says Jev cannot do. .gitignore gains a jev-judge allowlist block in the same shape as plan-integration (small Markdown/JSON only). The spike reads TYPESAFE_API_KEY from the environment; the key lives in an ignored .env.

…sure one judge Planning round for two work items the owner asked for after the Jev Stage 0 spike: (1) make the mentor agent actually write learning signals on every supported harness, proven by programmatic simulation of a learner; (2) one pre-registered, directional measurement of Jev as a filter over harness- proposed struggle candidates on the frozen 13-session gold. Brief, three seats (gpt-6-astra, qwen3-coder, grok-4.6 re-run), arbitration and the resulting plan are all on the record here. Why the plan differs from the brief the seats reviewed: verifying their claims against source overturned four of them and sharpened the rest. The largest: the gold labels (6c7176e, 2026-05-31) were authored against the pre-rebuild corpus, and 9 of the 10 gold sessions present in both databases now have DIFFERENT transcripts in the live file (the 2026-09-12 clean start stripped tool echoes; one Claude session 195 -> 49 messages). So the ruler is the cold archive, all 13, mode=ro -- not the live-union the brief proposed. Also verified: both DBs are schema 48 (Grok B4 falsified), claude_code is in STUDY_SOURCES (B3 falsified), eval_runner already opens read-only (B2 partly falsified), C1's recall varied 1/9..7/9 across the 83 ledger rows and was never run held-out (Astra F05 confirmed; corrects the coordinator's own "stuck at 1/9" claim), all six harness binaries are installed here (Astra F01 achievable), held-out has 2 negatives (0/2 FP has a 77.6% one-sided upper bound -- directional only). Decisions carried into the plan: MCP record_teachback with a validator shared with the CLI; writer set split W_auto (additive, pre-approved) vs W_srs (SRS mutators accept unverified card hashes -- stay prompt-per-call); one YAML trigger table projected to all six harnesses with CI byte-identity; definition of done states exactly which harnesses passed and counts no skip as a pass; Jev measured only as the delta between matched candidate lists with and without its filter; no adopt/reject clause; two branches, two PRs (deviates from the owner's "one branch" ask -- flagged for decision). Grok's first seat looped for 24,000 tokens announcing repo inspections it could not perform (kept as seat-grok-4.6.INVALID-tool-loop.md, truncated); the re-run with scripts/council/system-seat.md is the valid seat. .gitignore: learning-tier allow-list block, same shape as plan-integration. .pre-commit-config.yaml + .secrets.baseline: council manifest/receipt JSON under learning-tier excluded from the hex-entropy detector, same class and rationale as plan-integration (sha256 digests of committed public text). Baseline regenerated whole-repo with the pinned v1.5.0: results 72 -> 72, only the filter pattern changed.

openspec/ was ignored at .gitignore:644 while 78 files under it are tracked and load-bearing (the release gate reads openspec/changes, CI validates the specs). That is the tracked-but-ignored trap this file's own comment warns about: every new archive or spec file was invisible to `git status` and skipped by `git add -A`, so archives had to be force-added. Nothing else under openspec/ was being ignored (git status --ignored: only the archive created this morning), so removing the rule surfaces exactly the files that should have been visible all along.

…ed, deltas not merged The change has said `deferred:` in its .openspec.yaml since 2026-09-05 (the owner kept tmux as the production default; herdr stays an experimental opt-in, ghostty dev-only), but it sat un-archived with five open boxes, so openspec/changes held a live directory nobody was working on. Each of the five boxes now carries an explicit "closed 2026-09-23 - NOT built, deferred with the change" disposition naming its reopen condition (T1.2 dedicated TmuxBackend suite, T1.3 settings.py mention, T1.4 old-format round-trip test, T1.6 test_session_start mock target, T4.2 T6 herdr detach xfail), so the archive counts complete without claiming work that was never done. Archived with --skip-specs. Verified against the tree, not recalled: all eight MODIFIED requirement headers exist in no main spec (grep of openspec/specs for each header: no file), so they were modifications of requirements never written; the web-ui delta preserves the wterm selector e9cb565 removed and a bootstrap e4b17d4 replaced; the session-transports delta routes a ttyd fallback the PTY start path calls retired. The delta files stay in the archive as the record. Main specs byte-identical (sha256 of all 23 spec.md before/after). openspec/changes now holds only archive/. `openspec validate --archived --all`: the new archive passes (7 passed; the one failure is the July archive that fails identically on main). `--specs --all` 23/23. The release gate (validate_new_archives) will validate this archive at 0.5.1 and it passes today.

…history.progress's own Receipt now-rubric-2026-09-16, row 3c reading (e) (2026-09-21): the medium screen printed "last seen 8 day(s) ago" for a struggle planted three days before the frozen date. spaced_repetition_due() reads datetime.now(UTC) in history/progress.py; the guidance tests freeze decision.datetime only. The count drifts with the real date, and with it the due copy's score (100 + min(days, 30) + 35), so a pinned screen can re-rank on a later day. The test patches progress.datetime with a decoy (2031) so it fails on any wall-clock date unless the count comes from the engine's instant. Fails now: days_ago 1570 where 3 is planted.

…s own build_now_plan reads datetime.now(UTC) once; the due-progress collector passed nothing on and spaced_repetition_due() read a second clock inside history/progress.py. Under the guidance tests, which freeze decision.datetime only, the due copy of a struggle planted three days before the frozen date printed "last seen 8 day(s) ago" on 2026-09-21 and 10 on 2026-09-23 — and its score (100 + min(days, 30) + 35) drifted with it, so a pinned screen could re-rank on a later date. In production both reads were the same wall clock microseconds apart, so no learner saw the split (receipt now-rubric-2026-09-16, row 3c (e)). spaced_repetition_due gains a keyword-only `now` defaulting to the wall clock (recap, review CLI, planning evaluation and authoring are unchanged); _due_progress_candidates takes the engine's instant and hands it on; build_now_plan passes its one read. Deliberate test change: the six stubs of _due_progress_candidates in test_learning_decision.py and test_now_plan_guidance.py accept **_ so the seam's new keyword reaches them. One RED expectation corrected at GREEN: the primary's score is the collector's 138 plus PLAN_RELATED_BIAS (rule 5), asserted from the constant. The struggle collector's own today = datetime.now(UTC).date() is left as is: it lives in decision.py, inside the frozen module, so it has no gap. RED flipped; guidance + decision + history + recap + web-now + web-history 153 passed; ruff/format/pyright clean; golden ec451ce8 byte-identical; docs contract green. Receipt row 3c records the fix beside the gap.

… four boxes, all built openspec validate --archived --all has reported one failure on every run since the release guard was added: 2026-07-12-complete-e2e-harness-and- desktop-mcp with four open boxes, which the Justfile excused as "unticked tasks nobody has evidence to reconcile". The evidence exists; the notes were written before the same-day commits that built the items: 1.4 fake-agent PTY binary — tests/_fake_agent.py (reads the persona, asks the Socratic bank), test_fake_agent.py, spawned by the journey and body-double/ghostty e2e files. 4.1 journey through generation/review/session end — landed across test_representative_user_journey.py (fake-agent walk: start, turn, end, durable row) and test_journey_generate_review.py (phases 2-4). 4.4 session-end export asserted — study_sessions row after /api/session/end; card_reviews/review_sessions rows after review. 5.1 MCP get_lesson_tree/read_lesson/search_lessons — 9f09cf0, dated the same day as the "confirmed absent" note; tested in test_mcp_tools.py. The original notes are kept and each disposition is appended with its pointer, as the herdr archive does. openspec validate --archived --all: 8 passed, 0 failed.

…ly archive it cited is reconciled The Justfile comment and validate_new_archives docstring said the July archive "has unticked tasks nobody has evidence to reconcile". As of the previous commit it does not. The scoping to NEW archives stays — a historical archive nobody is working on must never re-fail a release — so the comments now state the design reason and record that the motivating case is closed.

…d, herdr archived as deferred, July archive reconciled, one-clock fix

Copilot

Copilot review overview

🟡 Changes recommended

Unresolved moderate findings remain in graphify exclusions, clock propagation, Jev model pinning, and lesson-template serialization.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 3 · 2

Open (5)

Thread engine time through all date-based collectors · New
Fail instead of falling back to an unpinned model · New
Preserve pre-scan exclusion before removing graphifyignore · New
Correct incorrect receipt character count · New
Update plan status to reflect resolved decision 5 · New

What changed in this PR

This housekeeping PR publishes the study-notes skill, records Jev planning evidence, reconciles OpenSpec archives, fixes frozen-clock handling, and cleans up graphify configuration.

Changes:

Adds study-notes documentation, templates, and navigation.
Archives and reconciles deferred OpenSpec work.
Threads the engine clock through due-progress calculations.
Updates planning, release checks, changelog, and repository metadata.

File	Summary / review note
`skills/studyloop-study-notes/SKILL.md`	Adds the study-notes workflow.
`skills/studyloop-study-notes/references/xtiles.md`	Documents xTiles presentation.
`skills/studyloop-study-notes/references/obsidian.md`	Documents Obsidian workflows.
`skills/studyloop-study-notes/references/evidence-and-updates.md`	Defines evidence and update rules.
`skills/studyloop-study-notes/assets/section-overview.md`	Adds the overview template.
`skills/studyloop-study-notes/assets/lesson.md`	Adds the lesson template. Moderate (1 vote): quote the sequence placeholder so values such as `001` retain their exact text.
`scripts/graphify_refine.py`	Updates graphify handling. Moderate (2 votes): preserve a pre-scan exclusion for the vendored browser tree before removing `.graphifyignore`.
`scripts/eval/jev_stage0_spike.py`	Adds the Jev access spike. Nit (1 vote): record the uncertain factual-error control as unsuccessful/attempted. Moderate (2 votes): do not retry without the pinned model.
`scripts/check-release-consistency.py`	Documents archive validation. Nit (1 vote): clarify that it runs `openspec validate --archived --all` while filtering newly added archives.
`packages/studyloop/tests/test_now_plan_guidance.py`	Adds frozen-clock regression coverage.
`packages/studyloop/tests/test_learning_decision.py`	Updates collector test stubs.
`packages/studyloop/src/studyloop/learning/decision.py`	Forwards the engine instant. Moderate (3 votes): thread the instant through all date-derived collectors, including energy-demand and struggle classification.
`packages/studyloop/src/studyloop/history/progress.py`	Supports an explicit scheduling instant.
`openspec/changes/archive/2026-09-23-herdr-ghostty-multiplexer-transport/tasks.md`	Archives deferred transport tasks. Nit (1 vote): accurately retain the unbuilt test gap and reopen condition.
`openspec/changes/archive/2026-09-23-herdr-ghostty-multiplexer-transport/specs/web-ui/spec.md`	Preserves the deferred web UI delta.
`openspec/changes/archive/2026-09-23-herdr-ghostty-multiplexer-transport/specs/session-transports/spec.md`	Preserves the deferred transport delta.
`openspec/changes/archive/2026-09-23-herdr-ghostty-multiplexer-transport/specs/live-session-orchestration/spec.md`	Preserves the deferred orchestration delta.
`openspec/changes/archive/2026-09-23-herdr-ghostty-multiplexer-transport/proposal.md`	Archives the deferred proposal.
`openspec/changes/archive/2026-09-23-herdr-ghostty-multiplexer-transport/design.md`	Archives the deferred design.
`openspec/changes/archive/2026-09-23-herdr-ghostty-multiplexer-transport/.openspec.yaml`	Records deferral metadata.
`openspec/changes/archive/2026-07-12-complete-e2e-harness-and-desktop-mcp/tasks.md`	Reconciles historical archive tasks.
`mkdocs.yml`	Publishes the study-notes guide in navigation.
`Justfile`	Updates release-check documentation. Nit (1 vote): describe the actual all-archive invocation and filtering.
`docs/study-notes-skill.md`	Adds the public skill guide.
`docs/architecture/learning-tier/plan-2026-09-19.md`	Adds the learning-tier plan. Nit (2 votes): mark decision 5 resolved and leave only decisions 1–4 pending.
`docs/architecture/learning-tier/council/spend-round1.json`	Records council usage.
`docs/architecture/learning-tier/council/seat-qwen3-coder.md`	Adds council review.
`docs/architecture/learning-tier/council/seat-openai.gpt-6-astra.md`	Adds council review.
`docs/architecture/learning-tier/council/seat-grok-4.6.INVALID-tool-loop.md`	Records the invalid council run.
`docs/architecture/learning-tier/council/plan-grok-rerun/seat-grok-4.6.md`	Adds the valid rerun review.
`docs/architecture/learning-tier/council/plan-grok-rerun/manifest.json`	Records rerun metadata.
`docs/architecture/learning-tier/council/brief-plan-2026-09-19.md`	Adds the council brief.
`docs/architecture/learning-tier/council/arbitration-plan-2026-09-19.md`	Records arbitration decisions.
`docs/architecture/jev-judge/receipts/stage0-receipt.json`	Adds raw Jev receipt data.
`docs/architecture/jev-judge/receipts/stage0-access-2026-09-19.md`	Documents the Jev spike. Nit (2 votes): correct the `state_chars` count to match the receipt.
`CHANGELOG.md`	Records the housekeeping changes.
`.secrets.baseline`	Regenerates secret-scan metadata.
`.pre-commit-config.yaml`	Excludes council receipt manifests.
`.graphifyignore`	Removes tracked graphify exclusions. Moderate (1 vote): retain an equivalent pre-scan ignore for the vendored browser tree.
`.graphify-labels.json`	Removes tracked curated labels.
`.gitignore`	Un-ignores OpenSpec and planning records.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

+    ``now`` is the engine's one clock read (``build_now_plan``): the collector
+    counts ``days_ago`` from it rather than from a second read inside
+    ``history.progress``, so the day count on a screen — and the score built
+    from it — cannot disagree with the rest of the plan (receipt
+    ``now-rubric-2026-09-16``, row 3c (e): a frozen engine printed a drifting
+    "last seen 8 day(s) ago" for a struggle planted three days back).


+    try:
+        resp = client.system_one(state=STATE, questions=QUESTIONS, model=MODEL)
+    except TypeError:
+        # SDK may not take model= on this call; fall back and record what answered.
+        resp = client.system_one(state=STATE, questions=QUESTIONS)


+Curated labels live in ``.graphify-labels.json`` at the repo root — a local,
+gitignored file (untracked since 2026-09-22); the script runs without it and
+applies no curated labels when it is absent.


+  the environment, never prints it). SDK `typesafe-sdk==0.7.0`, Python 3.12.8.
+- Model pinned to `jev-1.13.0` — never the `jev-latest` alias, because any threshold
+  tuned against an alias silently moves when the alias does (vendor's advice, adopted).
+- State: one synthetic teach-back (1,873 chars) — a networking-background learner


+**Status:** council-validated planning round (3 seats, all ACCEPT-WITH-CORRECTIONS; every BLOCKING/MAJOR
+finding verified against source — see `council/arbitration-plan-2026-09-19.md`). **Owner decisions
+outstanding:** §7. **Base:** `main` @ `4f8e3e0f`.


NetDevAutomate · 2026-09-23T07:45:54Z

Full suite on this head (`c1a28de1`)

uv run --group dev pytest -q -p no:cacheprovider -rfE on the macOS sandbox host: 7316 passed, 31 failed, 14 errors, 17 skipped in 16m07s.

Failing/erroring ids (45) compared against the committed environmental set in docs/architecture/plan-integration/receipts/full-suite-control-item4-2026-09-18.md (51 ids = 44 environmental + item 4's seven then-REDs):

run − environmental set = one id: packages/agent-session-tools/tests/test_sync_conversation_integrity.py::test_concatenated_remote_dump_with_existing_archive — pytest-timeout at 60 s inside subprocess.run(["sqlite3", "-bail", …]).communicate(). It fails identically on a clean main control worktree with the same venv, and this branch does not touch packages/agent-session-tools (git diff --stat main..HEAD -- packages/agent-session-tools is empty). Pre-existing on this host; not a regression; not fixed here — it needs its own look (the host sqlite3 CLI never returning to the pipe).
environmental set − run = the seven item-4 REDs, green since 2026-09-18, as expected.

So: zero regressions attributable to this branch; the 44 shared ids are the known sandbox set.

NetDevAutomate added 12 commits September 23, 2026 07:58

docs(changelog): [Unreleased] — housekeeping: openspec tree un-ignore…

c1a28de

…d, herdr archived as deferred, July archive reconciled, one-clock fix

Copilot AI lite review requested due to automatic review settings September 23, 2026 07:21

NetDevAutomate mentioned this pull request Sep 23, 2026

Learning-tier plan (jev-judge): four owner decisions before S1-0 starts #34

Closed

Copilot started reviewing on behalf of NetDevAutomate September 23, 2026 07:21 View session

Copilot AI reviewed Sep 23, 2026

View reviewed changes

NetDevAutomate merged commit c1a28de into main Sep 23, 2026
17 checks passed

NetDevAutomate deleted the chore/housekeeping-0.5.x branch September 23, 2026 08:11

Navigation Menu

Housekeeping: skill guide published, jev plan on main, openspec tree un-ignored and archived, one-clock fix - #35

Housekeeping: skill guide published, jev plan on main, openspec tree un-ignored and archived, one-clock fix#35
NetDevAutomate merged 12 commits into
mainfrom
chore/housekeeping-0.5.x

NetDevAutomate commented Sep 23, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

NetDevAutomate commented Sep 23, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

NetDevAutomate commented Sep 23, 2026

Summary

Verification (all on this head)

Owner steps after merge

Not in this PR, on purpose

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

NetDevAutomate commented Sep 23, 2026

Full suite on this head (c1a28de1)

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Full suite on this head (`c1a28de1`)