release: 0.10.4, and CI that tests what ships

[askalf]

What this changes

Two commits. Review them separately.

1. release: 0.10.4 (a22d6e0)

Bumps version to 0.10.4. When this merges, auto-release.yml tags v0.10.4, cuts the GitHub release from the CHANGELOG section and dispatches publish.yml.

• package.json / package-lock.json: 0.10.3 → 0.10.4
• CHANGELOG [0.10.4]: the Windows EPERM lock fix (lock: a guard mid-release reads as EPERM on Windows — treat it as contention #197), the --version fix below, and a note that this release carries the current README to npm and Glama
• Pinned version bumped in docs/claude-code.md, docs/commands.md and both issue templates
• README watch figure: 2026-09-25 run (314 plugins · 2,442 skills · 0 under review)

docker/ stays on 0.10.3 on purpose. It installs from npm, so dependabot's /docker entry bumps it once 0.10.4 is published.

2. CI hardening (ed18322)

I reviewed every workflow. All recent runs were green, and everything passes actionlint and shellcheck, so the gaps were in what CI never checked:

Workflow	Change	Why
ci.yml	New package job: installs the npm pack tarball into an empty project (ubuntu + windows, node 20), then runs the installed CLI, truecopy-mcp and the library	The test suite imports ./src straight from the checkout, so a broken published package was invisible. This job found the --version bug on its first run.
ci.yml	Node 24 added to the matrix	Active LTS, and what publish.yml already uses. I ran the suite on 24.21.0 locally and it passes.
ci.yml	New actionlint job, with the binary pinned by version and sha256	Lints every workflow, and every run: script through shellcheck
marketplace-watch.yml	Also runs on master pushes that touch acceptances or detection; runs are serialized	#207 merged 40 minutes after the run that flagged its skill, so the badge stayed red until I dispatched a run by hand. Every run also force-pushes watch, so two runs must never overlap.
publish.yml	After npm publish: install that version back from the registry, check --version, check it flags the poisoned demo, run npm audit signatures	A green publish only means npm accepted the upload. I confirmed audit signatures verifies both the registry signature and the provenance on 0.10.3.
truecopy-gate.yml	verify-attestation: 'true'	The self-gate now checks the Sigstore provenance of the tarball it installs. I confirmed v0.10.3 has an attestation.
codeql.yml	Adds the actions language	CodeQL also scans the workflows themselves
docker.yml	Also runs when truecopy.lock or demo/clean-mcp.json change	Both are copied into the image, and the image's smoke test verifies them
test/release-hygiene.test.mjs	Checks that the CHANGELOG section, lockfile version and documented pins all match package.json	Catches a version bump that leaves any of them behind. The test fails on a mismatched bump and passes on this tree.

Bug fix: truecopy --version printed the usage text and exited 2, even though both issue templates ask reporters for its output. --version, -v and version now print the version. Regression test added.

Why

This PR gets the lock fix onto npm, and it adds tests for the package users actually install, the publish step and the watch's behavior after a triage PR merges.

How it was verified

• npm test: 274 tests, 273 pass, 0 fail, 1 skipped (Node 22). Node 24.21.0 locally: 269/269 before the new tests.
• actionlint 1.7.12 with shellcheck 0.11.0: clean on all workflows
• I ran the package job's two scripts locally, taken straight from ci.yml. The installed CLI reports 0.10.4, the clean demo passes, the poisoned demo exits 1, verify passes, truecopy-mcp lists truecopy-verify, truecopy-status, and the library has 33 exports.
• npm audit signatures against the published @askalf/truecopy@0.10.3: 1 verified registry signature, 1 verified attestation
• Detection change? N/A
• Acceptance change? N/A

🤖 Generated with Claude Code

https://claude.ai/code/session_01XZEotnWUeFC5V4AHMKBM9W

[sprayberry-redline]

Automated review from the Sprayberry Labs fleet code reviewer.
Reviewed by the GPT gating lane (gating review).

Verdict: changes requested — the commit carries public AI attribution, which first-party strict mode treats as a blocking generated-writing tell. rule:reads-as-generated

Blocking — commit a22d6e0: public AI attribution

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

The sole commit adds a model-specific co-author trailer. The PR also exposes the same generated provenance in its branch/body, but the trailer alone is sufficient: it makes the release history read as generated and is prohibited by the first-party strict-mode public-text pass. Remove the model-specific attribution (and generated-session material from the public PR metadata) before submitting the release.

# Remove the model-specific co-author trailer from the commit message.

The version values are consistent across package.json, the lockfile, pinned documentation, templates, and the changelog. I also confirmed all required CI checks are passing at a22d6e0ff0d3a946080567f1f494e834cfe02006; I did not run the local suite.

[askalf]

Replaced by a clean PR with the same change as one commit, without the model attribution in the commit trailer and body.