[Enard52]

What

The four verified items from the 2026-09-21 doc audit in #261, all one-session work. Docs + a single manifest field; no code paths touched.

• mcpb/manifest.json — "license": "MIT" → "Apache-2.0". pyproject.toml (license = "Apache-2.0"), the LICENSE file (Apache text) and the README all already say Apache-2.0; the bundle manifest was the odd one out.
• docs/analysis/*.md — added a one-line point-in-time field report banner to each of the four dated narratives, so readers know the paths/versions quoted inside reflect the tree at write-time (e.g. the 2026-07-03 RM1PE report still cites drivers/pikvm.py for the GL driver, moved by Refactor: split GLKVMDriver into its own drivers/glkvm.py (separate from pikvm.py) #140). Files kept exactly where they are — wheels/ and llms.txt reference these paths.
• docs/CONTRIBUTING.md — added a Release checklist section (there was no release section at all) with an explicit "refresh or delete RESUME.md" step: RESUME.md goes stale every release yet CLAUDE.md tells every session to read it first.
• README.md — the "latest unreleased tree" install example recommended kvm-pilot[totp,ws], but ws is a declared no-op back-compat alias (websocket-client is now a base dependency, see pyproject.toml). Dropped to kvm-pilot[totp].

Deliberately out of scope

• Two more [...ws] occurrences I left for you to decide on, since docs: hygiene drive-bys — mcpb license field, analysis banners, RESUME.md release step, GL REST-block dedup #261 named only README:109:
  • docs/CONTRIBUTING.md:11 — pip install -e ".[dev,totp,ws]" (editable dev install)
  • src/kvm_pilot/skill/references/setup.md:33 — same kvm-pilot[totp,ws] @ git+… VCS form
    Say the word and I'll fold them into this PR.
• The GL REST-block / pip-install---pre dedup ("collapse restatements to a one-liner + link") — that's a multi-file prose rewrite with real judgment calls, so I kept it out of a hygiene drive-by. Happy to take it as a follow-up.

Verification

• mcpb/manifest.json still parses (json.loads OK); license now Apache-2.0.
• pytest tests/test_docs_guards.py tests/test_mcpb_bundle.py → 40 passed. _current_doc_files() treats docs/analysis/ as historical (never command-line-guarded), so the banners don't trip the install-consistency guard; the canonical pip install --pre kvm-pilot line in README is untouched.
• Diff is 7 files, +26 / −2.

_{Disclosure: drafted with AI tooling under my review; each change above was checked against the actual files and the two relevant test modules run locally.}

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

• Configuration used: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: d2f4afbe-7be3-49d0-94dc-5d66d8cdb6ea

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

[DustinTrap]

Thank you for taking this on, and for the care in it — the setup.md claim you caught in #277 ("deliberately picks up no pre-release") was wrong, exactly as you said, and the out-of-scope notes in #276 were right too.

An apology is owed: a maintainer pass over the open issues ran in parallel without first checking the PR queue, and its versions of the same fixes (#280 for #254; #283 for #261) were merged before yours were reviewed. That is on us, not on you.

Where your work shows up:

• The 1.0.0 release PR credits fix(health): emit INFO for firmware currency data gap, not silent None (#254) #275, docs: hygiene drive-by — mcpb license, analysis banners, release checklist, ws extra (#261) #276 and docs: one home for the --pre rationale and the GL 404 note; fix the skill's bare-install claim (#261) #277 by name in the CHANGELOG.
• Your docs: hygiene drive-by — mcpb license, analysis banners, release checklist, ws extra (#261) #276 out-of-scope items (.[dev,totp,ws] in CONTRIBUTING, the [totp,ws] VCS line in setup.md) are fixed in the release PR.
• The setup.md claim from docs: one home for the --pre rationale and the GL 404 note; fix the skill's bare-install claim (#261) #277 is gone: with 1.0.0 the install command drops --pre everywhere, so the whole pre-release note is retired rather than corrected.
• fix(health): emit INFO for firmware currency data gap, not silent None (#254) #275's shape (INFO for the data gap, no flash prompt) is what health: "Firmware currency unknown" instead of silence when the registry has no latest (#254) #280 does as well; health: "Firmware currency unknown" instead of silence when the registry has no latest (#254) #280 additionally reports an explicit OK when current and treats a device absent from the registry the same way.

I'm leaving these open for you to close, so the record stays yours. Hardware reports and any further doc audit are very welcome — next time the queue gets checked first.

[Enard52]

Thank you for the detailed write-up and for crediting #275, #276 and #277 by name in the 1.0.0 CHANGELOG - that is more than generous, and the record being kept is the part that mattered to me.

Since #280 and #283 already cover the same ground and these three branches now conflict with main, I am closing them to keep your queue clean, as you suggested. Noted on /take before starting - I will claim issues first next time.

Happy to pick up hardware reports or another doc audit whenever useful.