Promote the Write Guard's gh Target Reading and the Windows Bootstrap Fixes to Main

.agents/skills/pr-review-conduct/SKILL.md

@@ -12,7 +12,7 @@ description: >-
   nothing about whether the review covered the current head SHA, read the full diff, or left a
   suppressed low-confidence finding, which opens no thread at all, unanswered. Also triggers when
   a review loop looks stuck (no review landing, findings that keep reappearing) or when deciding a
-  finding is real, false, deferred, or a deliberate decline, or when a reviewer looks missing or
+  finding is real, false, deferred, or a deliberate decline, or when a review looks missing or
   skipped. This skill is the contract that `scripts/pr_review.py`, `drive-pr`, and
   `merge-and-release` implement: running the loop hands-off is `drive-pr` and merging main is
   `merge-and-release`, each winning for its own action while this skill still binds the gate.

.agents/skills/python-codestyle/SKILL.md

@@ -64,8 +64,10 @@ default baseline on first-party code (a repo may instead run mypy with its stric
 and keep pyright editor-only via Pylance, per the next paragraph): `[tool.pyright]`
 `strict = ["src"]`, or the integration package for a Home Assistant repo, with tests run in
 standard mode. pyright is the anchor because Pylance embeds it, so where CI runs pyright, the
-editor and the CLI/CI (`uv run pyright` in the uv form, `uvx pyright@latest` in the pip form,
-pointed at the environment in a declared directory) run the same engine and never disagree.
+editor and the CLI/CI run the same engine. A pyright the requirements pin can lag the one Pylance
+bundles, though. The CLI/CI form is `uv run pyright` in the uv form. In the pip form a declared
+directory runs the environment's own pyright, or `uvx pyright@latest` pointed at the environment
+where it installs none.
 The standalone `ms-pyright.pyright` extension stays in
 `unwantedRecommendations` because Pylance covers it. Relax strictness on third-party code only
 when a dependency has no usable types and no alternative (e.g. `pandas`): a targeted, commented
@@ -124,6 +126,7 @@ carries a `[project]` table it then installs the directory itself as editable, w
 uv venv --clear                           # creates .venv, replacing an existing one
 uv pip install -r requirements.txt -r requirements-dev.txt   # one -r per requirements*.txt
 uv pip install -e .                       # only where pyproject.toml has a [project] table
+# in a declared directory whose .venv installs ruff, .venv/bin/python -m ruff replaces uvx ruff@latest
 uvx ruff@latest format                    # auto-format
 uvx ruff@latest check --fix               # auto-fix lint
 uvx ruff@latest check                     # verify lint clean
@@ -134,10 +137,14 @@ uvx ruff@latest format --check            # verify format clean
 Its type checker runs against that environment. mypy runs as `.venv/bin/python -m mypy` where the
 environment installs it, and otherwise as `uvx mypy@latest --python-executable .venv/bin/python`,
 adding `--python-version` with the environment's version unless the mypy config pins one. pyright
-runs as `uvx pyright@latest --pythonpath .venv/bin/python`. Where the config sits in the project
-directory, CI runs the checker there with no path, as in the uv form. A declared subdirectory with
-no config of its own uses the repository root's instead. CI then runs the checker from the root
-with the directory as its path and the interpreter as `<dir>/.venv/bin/python`. Run it the same
+runs as `.venv/bin/python -m pyright --pythonpath .venv/bin/python` where the environment installs
+it, and otherwise as `uvx pyright@latest --pythonpath .venv/bin/python`. ruff follows the same
+precedence, as `.venv/bin/python -m ruff` where the environment installs it, in place of the
+`uvx ruff@latest` above. In a directory the validator's `python-directories` input declares, a tool
+the requirements pin is then the version CI runs. Where the config
+sits in the project directory, CI runs the checker there with no path, as in the uv form. A
+declared subdirectory with no config of its own uses the repository root's instead. CI then runs
+the checker from the root with the directory as its path and the interpreter as `<dir>/.venv/bin/python`. Run it the same
 way, as in `<dir>/.venv/bin/python -m mypy <dir>`. On Windows each interpreter path ends in
 `.venv\Scripts\python.exe` instead. Those ruff commands and that type checker are the pip form's
 clean-compile, run with pytest before committing.

.agents/skills/python-codestyle/references/code-style.md

@@ -87,6 +87,16 @@
 - **Don't use exceptions for expected control flow.** Exceptions are for unexpected states.
 - **Don't suppress errors silently** (`except Exception: pass`). Either handle the specific
   exception and document why it's safe, or let it propagate.
+- **Don't leave a text-mode `subprocess` call to the locale's encoding.** Every call that reads
+  or writes a pipe as text (`text=True`, `universal_newlines=True`, or an `errors` argument) names
+  the encoding its child uses on that pipe. Below Python 3.15, a call that names none encodes and
+  decodes with the locale encoding unless UTF-8 Mode is on. That encoding follows the host's
+  locale, usually UTF-8 on Linux and the ANSI code page on Windows. The omission therefore looks
+  correct on a UTF-8 host. On any other, a Windows host being the common case, it misreads or
+  rejects non-ASCII output from a child writing UTF-8. No static check in the ruff, pyright, or
+  mypy toolchain looks for the omission in a subprocess call. Naming the encoding settles how the
+  bytes are translated, not what a byte or character outside that encoding is worth. Each call
+  decides that with its own `errors` argument, and the strict default decides it by raising.
 
 <!-- External -->
 

.agents/skills/python-codestyle/references/profiles.md

@@ -41,12 +41,12 @@ configuration alone marks the lint-only one:
   `[project.optional-dependencies]` extra. The pip form is a `pyproject.toml` beside a
   `requirements*.txt`, installed with pip, whether or not it carries a `[project]` table. A
   committed `uv.lock` makes a directory the uv form even where a `requirements*.txt` sits beside it.
-  In the pip form CI builds the environment as `SKILL.md` "Local development loop" shows, runs
-  pytest from it as `.venv/bin/python -m pytest`, and runs ruff through `uvx`. In a directory
-  declared in the hub validator's `python-directories` input it runs the type checker through
-  `uvx` pointed at that environment, or runs mypy from the environment where it is installed
-  there, while the undeclared default root runs a bare `uvx <checker>@latest` with nothing
-  installed.
+  In the pip form CI builds the environment as `SKILL.md` "Local development loop" shows, and runs
+  pytest from it as `.venv/bin/python -m pytest`. A directory declared in the hub validator's
+  `python-directories` input runs ruff and the type checker from that environment where it
+  installs them. It otherwise runs them through `uvx`, with the checker pointed at that
+  environment. The undeclared default root runs `uvx ruff@latest` and a bare
+  `uvx <checker>@latest`, with nothing installed.
 - **Scripts** (the `lint-only` profile): stdlib-only utility scripts embedded in a non-Python repo
   (e.g. a Python tooling subtree of a `csharp` app). Run the tools with `uvx` (no project install,
   no lockfile): the `pyproject.toml` carries only tool config (`[tool.ruff]`, `[tool.mypy]`, and

.agents/skills/workflow-ci-contract/references/d-guarantees.md

@@ -15,7 +15,7 @@ The required behaviors, organized by domain. Each is a **MUST**, and its `Output
 - **D1.3 Smoke never publishes and never uploads.** Input: `smoke: true`. Output: full compile/lint/test, but no registry/image push, no release, and **no** artifact uploads (every `upload-artifact`, including any aggregation job, is gated on smoke being false, written `!inputs.smoke` at the workflow layer and `inputs.smoke != 'true'` in a composite action, whose inputs are strings). *Prevents: a PR publishing, and orphaned artifacts churning the storage quota.*
 - **D1.4 Workflow-file changes are not smoke-built.** Input: a PR changing only `.github/workflows/**`. Output: the paths-filter marks no target, so smoke-build skips. An inclusion list satisfying D1.1 reaches this by leaving workflow paths out of every target's entry. *Implication: a workflow-only change is not smoke-built, but actionlint still validates it in CI.*
 - **D1.5 One required aggregator gates merge.** Input: any PR. Output: a single aggregator job must **succeed**, run under `if: always()` so a failed or skipped dependency cannot skip the gate itself, `needs:` the validation job, and the `changes` and `smoke-build` jobs too wherever the repo has a smoke build, treat a **skipped** smoke build as pass, and **block** on `failure`/`cancelled`. Its name is ruleset-bound: the job `name:` and the ruleset `context:` are the same string and MUST be renamed together, never independently. *Prevents: a paths-filter error letting a target-changing PR merge unbuilt.*
-- **D1.6 Coverage is reported to Codecov (C# and Python).** Input: a C# or Python repo that has tests for that type. Output: the validation job runs those tests under coverage collection (`dotnet test --coverage --coverage-output-format cobertura --results-directory ./coverage`, leaving `--coverage-output` unset so each test project writes its own report rather than overwriting a shared one, `pytest --cov-report=xml` in a Python directory with its own `uv.lock` or `requirements*.txt`, whose own `pyproject.toml` selects what to measure, or `coverage run -m unittest discover -s tests` then `coverage xml` in a `lint-only` one) and a `codecov/codecov-action` step uploads the report, **best-effort** (`continue-on-error` and/or `fail_ci_if_error: false`, so a Codecov outage or an absent token never reds the gate). Each leg **fails its test step when the run wrote no report**, since the best-effort upload reads a missing report exactly as it reads a healthy run, and on the Python leg `--cov-report=xml` alone selects nothing to measure. The C# leg renames each report to `coverage-<guid>.cobertura.xml` before the upload step reads the directory, `codecov-cli`'s own finder not matching the default name, and a repo owning its validator rather than calling the hub's owes that rename itself. `CODECOV_TOKEN` lives in the repo's **actions** and **dependabot** secret stores, the second because a run triggered by a Dependabot pull request reads the Dependabot store and the upload would otherwise skip silently on every bot pull request. A caller reaching the reusable validator across repositories names the secret it passes (`secrets:` with `CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}`), on its pull request path and its publisher path alike, because `secrets: inherit` is documented for a caller in the same organization or enterprise, which a personal account is not. A call by local path stays inside one repository, where the caller's own store is the one the callee reads, so `secrets: inherit` is available there instead of naming each secret. The repo ships a **`codecov.yml`** setting the project and patch statuses to **`informational: true`** so a coverage delta never gates a pull request (a repo whose quality bar requires a threshold may turn that off), and excluding intentionally-untested, non-shipped code (an example or benchmark project) from the denominator via `ignore`. Coverage output is a build artifact, so `.gitignore` excludes it. The C# invocation runs under **Microsoft.Testing.Platform**, and the runner declaration, package references, and version floor an MTP-based test project needs are `CODESTYLE.md`'s .NET side. The Python invocation needs **`pytest-cov`** and a coverage selector, which are `CODESTYLE.md`'s Python side. A Python directory declared in the validator's `python-directories` input owes a `tests/` suite whatever its profile, so the validator fails one that has none. N/A for a repo carrying no tests for that type. *Prevents: coverage silently going unreported, and a coverage regression blocking an unrelated pull request.*
+- **D1.6 Coverage is reported to Codecov (C# and Python).** Input: a C# or Python repo that has tests for that type. Output: the validation job runs those tests under coverage collection (`dotnet test --coverage --coverage-output-format cobertura --results-directory ./coverage`, leaving `--coverage-output` unset so each test project writes its own report rather than overwriting a shared one, `pytest --cov-report=xml` in a Python directory with its own `uv.lock` or `requirements*.txt`, whose own `pyproject.toml` selects what to measure, or `coverage run -m unittest discover -s tests` then `coverage xml` in a `lint-only` one) and a `codecov/codecov-action` step uploads the report, **best-effort** (`continue-on-error` and/or `fail_ci_if_error: false`, so a Codecov outage or an absent token never reds the gate). Each leg **fails its test step when the run wrote no report**, since the best-effort upload reads a missing report exactly as it reads a healthy run, and on the Python leg `--cov-report=xml` alone selects nothing to measure. The C# leg renames each report to `coverage-<guid>.cobertura.xml` before the upload step reads the directory, `codecov-cli`'s own finder not matching the default name, and a repo owning its validator rather than calling the hub's owes that rename itself. `CODECOV_TOKEN` lives in the repo's **actions** and **dependabot** secret stores, the second because a run triggered by a Dependabot pull request reads the Dependabot store and the upload would otherwise skip silently on every bot pull request. A caller reaching the reusable validator across repositories names the secret it passes (`secrets:` with `CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}`), on its pull request path and its publisher path alike, because `secrets: inherit` is documented for a caller in the same organization or enterprise, which a personal account is not. A call by local path stays inside one repository, where the caller's own store is the one the callee reads, so `secrets: inherit` is available there instead of naming each secret. The repo ships a **`codecov.yml`** setting the project and patch statuses to **`informational: true`** so a coverage delta never gates a pull request (a repo whose quality bar requires a threshold may turn that off), and excluding intentionally-untested, non-shipped code (an example or benchmark project) from the denominator via `ignore`. Codecov also reads a **`.codecov.yml`** at the repository root, so that name serves too. Coverage output is a build artifact, so `.gitignore` excludes it. The C# invocation runs under **Microsoft.Testing.Platform**, and the runner declaration, package references, and version floor an MTP-based test project needs are `CODESTYLE.md`'s .NET side. The Python invocation needs **`pytest-cov`** and a coverage selector, which are `CODESTYLE.md`'s Python side. A Python directory declared in the validator's `python-directories` input owes a `tests/` suite whatever its profile, so the validator fails one that has none. N/A for a repo carrying no tests for that type. *Prevents: coverage silently going unreported, and a coverage regression blocking an unrelated pull request.*
 
 ### D2 - Input/State Validation at Entry
 

.claude-plugin/fleet-skills/.source-digests/pr-review-conduct

@@ -1 +1 @@
-5aaa06a65aadb606
+97815b2d2157b2a4

.claude-plugin/fleet-skills/.source-digests/python-codestyle

@@ -1 +1 @@
-7bbaad0be74556f3
+a54a634cf8810359

.claude-plugin/fleet-skills/.source-digests/workflow-ci-contract

@@ -1 +1 @@
-76921cd09044d019
+82e59e34aeb7fac2

.claude-plugin/fleet-skills/skills/pr-review-conduct/SKILL.md

@@ -12,7 +12,7 @@ description: >-
   nothing about whether the review covered the current head SHA, read the full diff, or left a
   suppressed low-confidence finding, which opens no thread at all, unanswered. Also triggers when
   a review loop looks stuck (no review landing, findings that keep reappearing) or when deciding a
-  finding is real, false, deferred, or a deliberate decline, or when a reviewer looks missing or
+  finding is real, false, deferred, or a deliberate decline, or when a review looks missing or
   skipped. This skill is the contract that `scripts/pr_review.py`, `drive-pr`, and
   `merge-and-release` implement: running the loop hands-off is `drive-pr` and merging main is
   `merge-and-release`, each winning for its own action while this skill still binds the gate.

.claude-plugin/fleet-skills/skills/python-codestyle/SKILL.md

@@ -64,8 +64,10 @@ default baseline on first-party code (a repo may instead run mypy with its stric
 and keep pyright editor-only via Pylance, per the next paragraph): `[tool.pyright]`
 `strict = ["src"]`, or the integration package for a Home Assistant repo, with tests run in
 standard mode. pyright is the anchor because Pylance embeds it, so where CI runs pyright, the
-editor and the CLI/CI (`uv run pyright` in the uv form, `uvx pyright@latest` in the pip form,
-pointed at the environment in a declared directory) run the same engine and never disagree.
+editor and the CLI/CI run the same engine. A pyright the requirements pin can lag the one Pylance
+bundles, though. The CLI/CI form is `uv run pyright` in the uv form. In the pip form a declared
+directory runs the environment's own pyright, or `uvx pyright@latest` pointed at the environment
+where it installs none.
 The standalone `ms-pyright.pyright` extension stays in
 `unwantedRecommendations` because Pylance covers it. Relax strictness on third-party code only
 when a dependency has no usable types and no alternative (e.g. `pandas`): a targeted, commented
@@ -124,6 +126,7 @@ carries a `[project]` table it then installs the directory itself as editable, w
 uv venv --clear                           # creates .venv, replacing an existing one
 uv pip install -r requirements.txt -r requirements-dev.txt   # one -r per requirements*.txt
 uv pip install -e .                       # only where pyproject.toml has a [project] table
+# in a declared directory whose .venv installs ruff, .venv/bin/python -m ruff replaces uvx ruff@latest
 uvx ruff@latest format                    # auto-format
 uvx ruff@latest check --fix               # auto-fix lint
 uvx ruff@latest check                     # verify lint clean
@@ -134,10 +137,14 @@ uvx ruff@latest format --check            # verify format clean
 Its type checker runs against that environment. mypy runs as `.venv/bin/python -m mypy` where the
 environment installs it, and otherwise as `uvx mypy@latest --python-executable .venv/bin/python`,
 adding `--python-version` with the environment's version unless the mypy config pins one. pyright
-runs as `uvx pyright@latest --pythonpath .venv/bin/python`. Where the config sits in the project
-directory, CI runs the checker there with no path, as in the uv form. A declared subdirectory with
-no config of its own uses the repository root's instead. CI then runs the checker from the root
-with the directory as its path and the interpreter as `<dir>/.venv/bin/python`. Run it the same
+runs as `.venv/bin/python -m pyright --pythonpath .venv/bin/python` where the environment installs
+it, and otherwise as `uvx pyright@latest --pythonpath .venv/bin/python`. ruff follows the same
+precedence, as `.venv/bin/python -m ruff` where the environment installs it, in place of the
+`uvx ruff@latest` above. In a directory the validator's `python-directories` input declares, a tool
+the requirements pin is then the version CI runs. Where the config
+sits in the project directory, CI runs the checker there with no path, as in the uv form. A
+declared subdirectory with no config of its own uses the repository root's instead. CI then runs
+the checker from the root with the directory as its path and the interpreter as `<dir>/.venv/bin/python`. Run it the same
 way, as in `<dir>/.venv/bin/python -m mypy <dir>`. On Windows each interpreter path ends in
 `.venv\Scripts\python.exe` instead. Those ruff commands and that type checker are the pip form's
 clean-compile, run with pytest before committing.