release-readiness: weekly go/no-go

[github-actions]

release-readiness: weekly go/no-go — 2026-07-16

Release Readiness — Core

Verdict: ✅ READY to cut v3.6.0

There's one real feat sitting unreleased, the tree on main is green, versions are coherent, and there are no open blockers.

---

What would ship

One feature — nvim editor quality-of-life (feat(nvim)): five behaviours cherry-picked from an external (voidarc) config, each beating or filling a gap in ours:

1. i/a/A on a blank line auto-indents via the black-hole register (guarded on count == 0, so 3i/10a keep native behaviour);
2. async git fetch on VimEnter that toasts when upstream is ahead of HEAD (pinned to startup cwd, argv-form/Windows-portable, only after a successful fetch);
3. vimade dims inactive windows + cursorline shown only in the active window;
4. macro-recording indicator in the lualine mode block;
5. exrc for project-local config.

Both the git-fetch toast and exrc are gated on DOTFILES_OFFLINE (inert on engagement boxes), and exrc additionally leans on Neovim's vim.secure trust prompt. Includes the PR-review follow-ups (7a49d37, 0ff5f1f).

Diff: 7 files, +143 lines — all under nvim/ plus the CHANGELOG.md entry and a lazy-lock.json bump.

Proposed version + why

v3.6.0 (minor). The unreleased section is a single feat(nvim) with no breaking change → SemVer minor. core.version (3.5.2) matches the latest tag (v3.5.2) and the newest dated CHANGELOG heading — fully coherent, so release.sh will cleanly promote [Unreleased] → ## [v3.6.0] - 2026-07-16 and open a fresh section.

Blockers / pre-flight

None hard. Signals gathered:

• Audit / tree — ✅ green. ci passed on HEAD cf612a3 (PR feat(nvim): adopt five QoL tweaks from the voidarc config #228 merge). (A claude-routines dispatch — this run — is in progress; not a gate.)
• Open PRs — none open, so no freshness-triage Hold verdicts and no security bump waiting to ride the release.
• Pin freshness — ✅ all 7 zsh plugin pins current (update-plugins.sh --check). The nvim lock check needs an nvim binary (absent in this CI env) — advisory only; the lazy-lock.json was already re-pinned in this feature branch.
• Fleet drift — advisory; can't be verified from here (no OS repos checked out, fleet-drift.sh degrades gracefully). Every checked-out repo reports on-Core; the fan-out happens after the tag via make sync.

Next command

make release VERSION=3.6.0

Then follow RELEASE-RUNBOOK.md for the tag + OS-repo rollout (make sync off the green tag). This routine reports only — I haven't touched core.version, CHANGELOG.md, or run any release script.

Filed by the claude-routines workflow. Report-first: review and act — nothing was changed.

[github-actions]

release-readiness: weekly go/no-go — 2026-07-17

Verdict: HOLD — nothing to ship yet.

[Unreleased] is empty, and everything since v3.6.0 is either repo-meta docs or a routine lock bump. There's no user-visible Core work to justify a release.

What would ship

Nothing releasable. The 7 commits since v3.6.0:

Commit	Type	Vendored Core?
cde9ed9, 79354e1, 6a1799d	docs(runbook) → RELEASE-RUNBOOK.md	No — dev tooling
7c7fc2a	docs(strategy) → RELEASE-STRATEGY.md	No — dev tooling
3de82d4	chore(deps) → nvim/lazy-lock.json	Yes, but a freshness-bot lock refresh — trivial, no CHANGELOG entry, rides the next real release

The four docs commits touch only repo-meta files (not the vendored core/ surface). The single Core-file change is a routine nvim plugin-lock bump — the kind that settles into whatever release ships next, not a reason to cut one.

Version coherence — clean

• core.version = 3.6.0
• latest tag = v3.6.0
• latest CHANGELOG.md heading = [v3.6.0] - 2026-07-16

All three line up, with an empty [Unreleased] below. Nothing to promote.

Tree / fleet state

• Audit: green — the ci workflow on the current main HEAD (c61c47c) is success. (claude-routines still in-progress, but that's not the gate.)
• Fleet: not surveyed — moot with no release pending. It'll matter at the next real cut.

Next command

No release. When genuine feat/fix work lands and populates [Unreleased], re-run /release-readiness. Given the current base is v3.6.0, the likely next cut would be v3.6.1 (patch) if only fixes accrue, or v3.7.0 if a feat lands — but that's contingent on work that doesn't exist yet.

If you want the pending nvim lock bump to not sit unreleased indefinitely, the move is to let it accompany the next substantive change rather than cut a release for it alone.

Filed by the claude-routines workflow. Report-first: review and act — nothing was changed.

[github-actions]

release-readiness: weekly go/no-go — 2026-07-17

I have all the signals except the two network-gated advisory checks (fleet-drift, plugin freshness), which the sandbox blocks. Here's the verdict.

---

Verdict: ✅ READY to cut v3.7.0

The tree is green on main HEAD (c610e78), [Unreleased] holds a substantial, coherent body of work, and there are no open blockers. Version state is fully aligned (core.version 3.6.1 = tag v3.6.1 = latest dated CHANGELOG heading).

What would ship

A cross-fleet robustness + prompt/editor-capability release, grouped:

• Neovim clipboard reaches headless/remote boxes — a gated OSC52 last-resort provider closes the "yank does nothing over tmux/SSH" gap (never overrides a real provider); Windows paste uses Get-Clipboard -Raw so multi-line pastes stop CRLF-mangling.
• Windows-correct nvim toolchain — debugpy/venv resolve to Scripts\python.exe; LuaSnip skips its jsregexp build with no toolchain; LSP capability fetch is pcall-guarded; emmet_ls attaches to html.
• Starship prompt gains indicators — venv ($virtualenv), $package, git_metrics, WSL badge, an ENGAGEMENT/$env_var slot, and an opt-in docker_version custom (off by default).
• Hot-path performance — git_main_branch reads origin/HEAD in one call (was up to 18 show-ref subprocesses), git_current_branch uses --show-current, _core_suggest scores fork-free, pullall tallies in one awk pass, plus a dead tmux status-left branch and redundant zsh history setopts removed.
• Security — session-start hook now verifies pinned SHA-256 before extracting (fail-closed, no more curl | tar); claude-routines run least-privilege --allowedTools (closes a prompt-injection → OAuth-token exfil path); HISTORY_IGNORE covers --flag=value and env-assignment secret shapes.
• Fixed — :Cheatsheet moves off the deprecated nvim_buf_add_highlight to nvim_buf_set_extmark.

Proposed version + why

v3.7.0 (MINOR). The unreleased set includes feat-level capability additions — new starship indicators and a new nvim OSC52 clipboard backend (the feat(nvim,starship) commit 15f1155) — with no breaking changes. New features → minor bump. (Precedent: v3.6.0 was a single nvim feat → minor; the fix/refactor/security/docs-only releases before it were patches.)

Blockers / pre-flight

None hard. Two advisory checks I couldn't run here (network-gated in this sandbox) — run them before the cut, in runbook order:

• Fleet drift — confirm the OS repos aren't already lagging on a way that should settle first:

  bash scripts/fleet-drift.sh
  

• Pin freshness (zsh plugins; nvim via open PR, not a local rebuild):

  bash scripts/update-plugins.sh --check
  gh pr list --label automation/freshness-nvim-plugins
  

No open PRs at all right now → no freshness-triage Hold verdicts and no pending drift/security bump that ought to ride this release. CI on HEAD: ci ✅, parity-check ✅ (claude-routines in-progress is this run).

Next command

make release VERSION=3.7.0

Then drive RELEASE-RUNBOOK.md from there (this gate reports only — I haven't touched core.version, CHANGELOG.md, or any tag).

Filed by the claude-routines workflow. Report-first: review and act — nothing was changed.

[github-actions]

release-readiness: weekly go/no-go — 2026-07-19

These two advisory checks (fleet drift, zsh pin freshness) are blocked behind approval in this session, so I'll surface them as unverified pre-flight items with the commands to run. I have everything else needed for the verdict.

---

VERDICT: READY to cut v3.8.0 ✅

There's a coherent, shippable body of unreleased Neovim work; CI is green on main HEAD (b8c9968); versions line up; no open PRs or blockers. Two fleet-hygiene checks couldn't run in this session (approval-gated) — advisory, not blockers.

What would ship

All unreleased work is a Neovim UI + LSP polish pass (from feat(nvim): NvChad-styled UI refactor + its review-fix follow-up):

• NvChad-styled UI — hand-built palette-aware lualine statusline (mode/location pills, git/cwd blocks) and bufferline highlights (active buffer lifts with accent underline); both pcall-guarded with fallback on a fresh box.
• Unified LSP card style — hover / signature-help / diagnostic-float share one padded rounded border with size caps; signature help now auto-pops inside function args (CursorHoldI, gated on server support).
• which-key + <leader>? cheatsheet restyled to mirror NvChad's keymap visualizer (palette-aware, with fallback).
• Fixes — taplo TOML root detection (real manifest names, not the unsupported *.toml glob); <leader>oi organize-imports guard (skips servers that can't do it, drops the racy 50ms timer); cursor-restore skips gitcommit/gitrebase buffers; folding single-owner via nvim-ufo.
• Internal — shared utils.lsp.with_snippets helper; augroup { clear = true } standardization; stale-comment fix.

Proposed version + why → v3.8.0 (minor)

• core.version = 3.7.0, latest tag = v3.7.0, CHANGELOG [v3.7.0] dated — all coherent, [Unreleased] sits cleanly on top.
• Commits since v3.7.0 are feat(nvim): … + fix(nvim): …. A feat with no breaking change → minor bump: 3.7.0 → 3.8.0. (All changes are additive/behavioral within nvim; nothing breaking.)

Blockers / pre-flight

• None hard. CI ci on HEAD b8c9968 = ✅ success; sync-fanout succeeded after the v3.7.0 release; no open PRs in this repo.
• Advisory — not verified this run (approval-gated, run before you fan out):
  • Fleet on latest Core? → bash scripts/fleet-drift.sh
  • zsh pins fresh? → bash scripts/update-plugins.sh --check
  • nvim freshness PR? → none open in dotfiles-core (gh pr list = empty).

Next command

make release VERSION=3.8.0

Then follow RELEASE-RUNBOOK.md for the tag, OS-repo rollout, and htpx. (Clear the two advisory checks above first if you want a clean fan-out.)

Report only — I did not touch core.version, CHANGELOG.md, or run any release script.

Filed by the claude-routines workflow. Report-first: review and act — nothing was changed.

[github-actions]

release-readiness: weekly go/no-go — 2026-07-19

I have everything needed. Here's the readiness verdict.

---

✅ READY to cut v3.9.0

The tree is green, version metadata is coherent, there's a real feature's worth of unreleased work, and there are no open blockers.

What would ship

All unreleased work is Neovim — the NvChad-hybrid UI refactor (PR #244) plus its follow-up fixes:

• Added — shared utils/palette.lua (single tokyonight palette source), scroll-% statusline indicator, NvChad-style inline LSP renamer (<leader>rn), colorify-style inline colour highlighter.
• Changed — statusline/tabline go hybrid-NvChad (opaque bufferline blocks, colored blink.cmp kind-icons, opaque cheatsheet card); signature help now owned by blink.cmp; Mason install manifest moved to its own VeryLazy module so fresh boxes provision near startup.
• Removed — trimmed the unused DAP debugger stack, neotest, dadbod, incline, aerial, spectre, git-conflict, mini.indentscope (16 lock entries).
• Fixed — heavy linters restricted to BufWritePost; bash-LS SC1071 phantom diagnostic on zsh silenced; format-on-save mini.trailspace pcall-guarded.

Proposed version + why

v3.9.0 (minor). Commits since v3.8.0 are 1 feat(nvim) + 2 fix(nvim), no !/breaking. A feat with no breaking change → minor bump. core.version (3.8.0) and the latest tag (v3.8.0, 2026-07-18) are aligned, so release.sh will cleanly promote [Unreleased] → ## [v3.9.0].

Blockers / pre-flight

None hard. Two advisory checks I couldn't run here (both need network approval) — worth a glance but not gating:

• Fleet drift — sync-fanout already ran green on the v3.8.0 merge (da15412, 2026-07-19 02:30 UTC), so the fleet is on the latest Core. Confirm nothing drifted since: bash scripts/fleet-drift.sh
• Pin freshness — no open freshness-nvim-plugins or zsh-plugin PRs (gh pr list is empty), so no stale-pin PR wants to ride this release. Optionally: bash scripts/update-plugins.sh --check
• Open blockers — none. No open PRs, no Hold verdicts; latest ci on main HEAD (a5eb9cb) is success.

Next command

make release VERSION=3.9.0

Then drive RELEASE-RUNBOOK.md from there. One note for the runbook: this is an all-Neovim release, so the fan-out will refresh each OS repo's core/nvim and OS repos that use nvim-sync.ps1 (Windows) will want their nvim tree re-synced too.

Filed by the claude-routines workflow. Report-first: review and act — nothing was changed.

[github-actions]

release-readiness: weekly go/no-go — 2026-07-20

I have what I need. Fleet-drift and pin-freshness need approval/network (advisory only) — but note both freshness bumps are already merged into HEAD. Here's the verdict.

---

🟢 READY to cut v3.10.0

The tree is green, there's a coherent body of unreleased work, and it carries new features (not just fixes) — a clean minor release.

What would ship

This is a Neovim-focused release (18 commits since v3.9.0, all under nvim). The story, grouped:

• Added — two features
  • Python debugging via nvim-dap + nvim-dap-python under a new <leader>d group — lazy on debug keymaps/commands, zero startup cost, adapter resolved debugpy→uv→python3. Revives the previously-dead rustaceanvim DAP block.
  • Statusline venv block — shows the active uv/.venv name in Python buffers, memoised per-buffer.
• Perf — startup & file-open wins
  • File-plugins deferred to User FilePost after UI ready: opening a file 165.9ms → 99.0ms (-40%).
  • lazy.nvim now defaults specs to lazy = true (regression net); SchemaStore deferred to client start; treesitter parser set cached.
  • Stopped sourcing matchparen/rplugin/spellfile, dropped showmatch (per-keystroke insert-mode cost).
• Fixed — six correctness bugs
  • Statusline blanking on NvimTree focus; visual-mode git staging silently staging whole hunk; SchemaStore never reaching jsonls/yamlls; gr* keymap cleanup deleting nothing; node/python3 providers disabled → checkhealth now 0 errors/0 warnings; gsn phantom mapping.
• Refactor — LSP server modules become plain data tables; one server list drives registration.

Proposed version + why

v3.10.0 (minor). Commit types since v3.9.0: 2 feat, 3 perf, 6 fix, 1 refactor, chore/docs. Two feat(nvim) entries and no breaking-change markers (!) → minor bump. (The docs(release) "v4 loader overhaul" commits are a draft proposal only — nothing breaking has landed in [Unreleased], so they don't force a major.)

Blockers / pre-flight

• None hard. CI on main is green (ci for merged chore(nvim): close cheatsheet gaps, name which-key prefixes, drop dead config #259 = success), no open PRs, no Hold verdicts.
• Version coherence ✓ — core.version (3.9.0) == last tag (v3.9.0); release.sh will promote [Unreleased] → ## [v3.10.0] and bump core.version.
• Advisory (couldn't run — needed approval/network):
  • fleet-drift.sh — worth a glance so you know how far behind the OS repos are before fanning out, but not a blocker.
  • update-plugins.sh --check — however, both freshness bumps (f162bd6 nvim lock, 045ced1 zsh pins) are already merged into HEAD, so pins are current as of this release point.

Next command

make release VERSION=3.10.0

Then follow RELEASE-RUNBOOK.md for the tag, OS-repo rollout, and htpx. (Optionally eyeball bash scripts/fleet-drift.sh first so the post-release make sync fan-out has no surprises.)

Filed by the claude-routines workflow. Report-first: review and act — nothing was changed.