Record why the host tooling carries no linter category, or decide that it should

[ptr727]

Neither host-setup/linux/install-tools.sh nor the new host-setup/windows/install-tools.ps1 manages markdownlint, cspell, actionlint, editorconfig-checker, shellcheck, PSScriptAnalyzer or ruff, and nothing in the tree states that as a decision. A reader can only reach it by inference.

The reasoning is sound and currently lives nowhere durable. Every one of those runs as a pinned container image, or through uvx for the Python toolchain, which is what keeps a local run and CI the same check: the image tag fixes the version. Installing native copies through apt or winget would put a second, unpinned version of each on the host, and a local run would then differ from CI, which is the exact property the pinned images exist to guarantee. So the only host requirements the whole linter surface creates are docker and uv, and docs/host-setup.md already carries both.

host-setup/windows/README.md now states this under "Why There Is No Linter Category", but it states it for Windows only. The question is whether it belongs in docs/host-setup.md as a fleet fact instead, so the Linux tooling is covered by the same sentence rather than by silence, and so a future contributor adding a "linters" category to install-tools.sh finds the answer before writing it rather than after.

Either outcome closes this. What should not persist is the current state, where the absence is correct and undocumented.

Raised while building the Windows host tooling, where the same absence had to be justified from scratch because nothing recorded the existing justification.

[ptr727]

Addressed by #710 (17003e6 on develop, promoted to main in #718 at 0e84805), which added host-setup/linux/README.md carrying its own "Why There Is No Linter Category" section.

Verified in the tree at aa7bf78. That section states the decision fleet-wide rather than per platform, in its first two sentences: "Neither this tooling nor its Windows sibling installs markdownlint, cspell, actionlint, editorconfig-checker, shellcheck, PSScriptAnalyzer or ruff. That is a decision rather than a gap." It then carries the same reasoning this issue recorded (pinned image or uvx, keeping a local run and CI the same check, a native install adding a second unpinned version, docker and uv as the only host requirements that follow) and cross-links the Windows README so the two sides state one decision.

So the Linux tooling is no longer covered by silence, which is what this issue said must not persist.

Closing on the "either outcome" clause rather than on an exact match: the resolution puts the fleet fact in the Linux README rather than in docs/host-setup.md. If centralizing it there is what you wanted, this is worth reopening, and it is a move rather than new writing.