Promote Develop: Split the Over-Length Comment Sentences and Make sentence-length a Default Check

.agents/skills/branching-and-release-model/SKILL.md

@@ -105,9 +105,8 @@ rather than guessing from the repo's contents.
   build height. That height is both wrong (the real height differs) and a maintenance trap.
   "Correcting" `1.0` to `1.0.0` is a defect.
 - **A no-op publish (unchanged NBGV `SemVer2`) on a schedule or push trigger re-pushes nothing to
-  any target keyed on the version string.** Docker is the exception: it always re-pushes (a dispatch
-  refreshes the release instead of skipping) to pick up upstream base-image
-  refreshes. Full guarantee and the `version.json` `pathFilters` boundary:
+  any target keyed on the version string.** Docker is the exception: it always re-pushes to pick up
+  upstream base-image refreshes. Full guarantee and the `version.json` `pathFilters` boundary:
   `references/release-publish-mechanics.md`.
 - **A package push can fail after the release is already cut**, since it runs after the release
   task and no gate covers it. A full re-run is always available inside its bounded

.agents/skills/comment-and-doc-style/SKILL.md

@@ -131,9 +131,8 @@ incrementally, and this section names each check as it ships.
 
 - **Short sentences: at most 25 words in one sentence**, ASD-STE100's descriptive cap, checked by
   the `sentence-length` rule in `prose_lint.py`. The rule reads Markdown prose, code comments, and
-  docstrings. The check is opt-in like `sentence-split`, because the existing corpus predates the
-  cap and a default gate would fail whole files nobody is editing. Write new prose under the cap,
-  and scope a run to a change with `--check sentence-length --diff <base>`.
+  docstrings. It runs by default, so a bare run checks it. A run with `--diff <base>` checks each
+  sentence a changed line belongs to, in full. Write new prose under the cap.
 - **One instruction per sentence.** A procedure step states one action, and a second action is a
   second step. No pattern reaches this, so it is authoring discipline with no check.
 - **Active voice, imperative mood for procedure steps.** Write "run the gate", never "the gate

.agents/skills/local-strict-review/SKILL.md

@@ -45,7 +45,7 @@ Task: adversarial review of this branch's diff against its merge-base with its t
   read full surrounding files where the diff hunks alone do not give enough context.
   Run `python3 <hub-checkout>/scripts/prose_lint.py --diff "$(git merge-base refs/remotes/origin/<target> HEAD)"`.
   Add `--allow-comments` where the change carries, or will open carrying, the `comments` label.
-  Then run the same command again with `--check sentence-length` added, and report each output as findings.
+  Report its output as findings.
 Worktree: <worktree>, the checkout holding the branch.
   Start every command with `cd <worktree> &&`, since a harness can reset the working directory between calls.
   Read each listed file at its path under `<worktree>`.

.claude-plugin/fleet-skills/.source-digests/branching-and-release-model

@@ -1 +1 @@
-ff60e0fba2b7e0c4
+6d8df70427b01dab

.claude-plugin/fleet-skills/.source-digests/comment-and-doc-style

@@ -1 +1 @@
-b8e088d656d9ff94
+846e20f174fbe3a7

.claude-plugin/fleet-skills/.source-digests/local-strict-review

@@ -1 +1 @@
-c5ee59056c13d054
+8303ad535ff03847

.claude-plugin/fleet-skills/skills/branching-and-release-model/SKILL.md

@@ -105,9 +105,8 @@ rather than guessing from the repo's contents.
   build height. That height is both wrong (the real height differs) and a maintenance trap.
   "Correcting" `1.0` to `1.0.0` is a defect.
 - **A no-op publish (unchanged NBGV `SemVer2`) on a schedule or push trigger re-pushes nothing to
-  any target keyed on the version string.** Docker is the exception: it always re-pushes (a dispatch
-  refreshes the release instead of skipping) to pick up upstream base-image
-  refreshes. Full guarantee and the `version.json` `pathFilters` boundary:
+  any target keyed on the version string.** Docker is the exception: it always re-pushes to pick up
+  upstream base-image refreshes. Full guarantee and the `version.json` `pathFilters` boundary:
   `references/release-publish-mechanics.md`.
 - **A package push can fail after the release is already cut**, since it runs after the release
   task and no gate covers it. A full re-run is always available inside its bounded

.claude-plugin/fleet-skills/skills/comment-and-doc-style/SKILL.md

@@ -131,9 +131,8 @@ incrementally, and this section names each check as it ships.
 
 - **Short sentences: at most 25 words in one sentence**, ASD-STE100's descriptive cap, checked by
   the `sentence-length` rule in `prose_lint.py`. The rule reads Markdown prose, code comments, and
-  docstrings. The check is opt-in like `sentence-split`, because the existing corpus predates the
-  cap and a default gate would fail whole files nobody is editing. Write new prose under the cap,
-  and scope a run to a change with `--check sentence-length --diff <base>`.
+  docstrings. It runs by default, so a bare run checks it. A run with `--diff <base>` checks each
+  sentence a changed line belongs to, in full. Write new prose under the cap.
 - **One instruction per sentence.** A procedure step states one action, and a second action is a
   second step. No pattern reaches this, so it is authoring discipline with no check.
 - **Active voice, imperative mood for procedure steps.** Write "run the gate", never "the gate

.claude-plugin/fleet-skills/skills/local-strict-review/SKILL.md

@@ -45,7 +45,7 @@ Task: adversarial review of this branch's diff against its merge-base with its t
   read full surrounding files where the diff hunks alone do not give enough context.
   Run `python3 <hub-checkout>/scripts/prose_lint.py --diff "$(git merge-base refs/remotes/origin/<target> HEAD)"`.
   Add `--allow-comments` where the change carries, or will open carrying, the `comments` label.
-  Then run the same command again with `--check sentence-length` added, and report each output as findings.
+  Report its output as findings.
 Worktree: <worktree>, the checkout holding the branch.
   Start every command with `cd <worktree> &&`, since a harness can reset the working directory between calls.
   Read each listed file at its path under `<worktree>`.

.coderabbit.yaml

@@ -1,16 +1,20 @@
-# The two trees below are what scripts/build_dist.py generates from .agents/skills/ and CI holds current, so a finding in either belongs at its source and a review of every copy is one finding three times.
+# The two trees below are what scripts/build_dist.py generates from .agents/skills/ and CI holds current.
+# A finding in either therefore belongs at its source, and a review of every copy is one finding three times.
 reviews:
-  # Auto review covers only the default branch unless a base is listed here, each entry a regex, and the default branch stays included whatever is listed.
+  # Auto review covers only the default branch unless a base is listed here, each entry a regex.
+  # The default branch stays included whatever is listed.
   auto_review:
     enabled: true
     base_branches:
       - "^develop$"
-    # A fleet pull request routinely passes five pushes before it merges, and the default pauses after five reviewed commits, which reads as a reviewer that stopped.
+    # A fleet pull request routinely passes five pushes before it merges.
+    # The default pauses after five reviewed commits, which reads as a reviewer that stopped.
     auto_pause_after_reviewed_commits: 0
   path_filters:
     - "!.github/skills/**"
     - "!.claude-plugin/fleet-skills/**"
-  # Canonical prose is linted for style in CI and read whole by the local review pass, so a review comment on it earns its place only by naming something false or unfollowable.
+  # Canonical prose is linted for style in CI and read whole by the local review pass.
+  # A review comment on it therefore earns its place only by naming something false or unfollowable.
   path_instructions:
     - path: "**/*.md"
       instructions: |
@@ -21,7 +25,8 @@ reviews:
   suggested_labels: false
   suggested_reviewers: false
   in_progress_fortune: false
-  # CI runs these four linters and fails the pull request on them, so a review comment from the same tool is a second copy of a red check.
+  # CI runs these four linters and fails the pull request on them.
+  # A review comment from the same tool is therefore a second copy of a red check.
   tools:
     markdownlint:
       enabled: false

.github/actions/nuget-build-default/action.yml

@@ -76,7 +76,8 @@ runs:
         path: |
           ${{ runner.temp }}/publish/*.nupkg
           ${{ runner.temp }}/publish/*.snupkg
-        # A project that packs nothing would otherwise only warn, create no artifact at all, cut a release, and fail one job later on a download that cannot say why.
+        # A project that packs nothing would otherwise only warn and create no artifact at all.
+        # It would still cut a release, and fail one job later on a download that cannot say why.
         if-no-files-found: error
         # Consumed within the run by the caller stub's publish-nuget job, so minimize artifact storage.
         retention-days: 1

.github/actions/prose-gate/action.yml

@@ -26,12 +26,17 @@ runs:
       id: gate
       shell: bash
       env:
-        # A caller reaches this gate at whatever hub commit its workflow pins, and only the action knows which one that is, since a checked-out action carries no history to read it from.
+        # A caller reaches this gate at whatever hub commit its workflow pins.
+        # Only the action knows which commit that is, since a checked-out action carries no history to read it from.
         # Unnamed, a finding from a stale pin reproduces against no local run and reads as a gate defect rather than as a version gap.
-        # Both the context and the default environment variable are read, because the context's value for a repository-relative reference is undocumented and the two are populated by different paths.
-        # Either being empty leaves the script to fall back to its own checkout, which is the correct answer whenever the action was not fetched from a pin.
-        # This repository calls the reusable workflow by a workspace-relative path, so its own runs may well take that fallback, and a downstream repository calling it at a pin is the case these exist for.
-        # What each form populates is read off a real run rather than predicted here, since predicting it is what a verdict naming its own copy exists to stop anyone having to do.
+        # Both the context and the default environment variable are read.
+        # The reasons are that the context's value for a repository-relative reference is undocumented, and that the two are populated by different paths.
+        # Either being empty leaves the script to fall back to its own checkout.
+        # That is the correct answer whenever the action was not fetched from a pin.
+        # This repository calls the reusable workflow by a workspace-relative path, so its own runs may well take that fallback.
+        # A downstream repository calling it at a pin is the case these exist for.
+        # What each form populates is read off a real run rather than predicted here.
+        # This is because predicting it is what a verdict naming its own copy exists to stop anyone having to do.
         ACTION_REPOSITORY: ${{ github.action_repository }}
         ACTION_REF: ${{ github.action_ref }}
       run: |
@@ -62,7 +67,8 @@ runs:
         # Check the base resolves before scanning, so an empty or absent ref fails naming itself.
         # Unresolvable, the run would report the repository's whole backlog against this change.
         # A shallow checkout is the usual cause, so the caller fetches full history.
-        # A base that is a freshly created merge commit can still lag GitHub's own replication, so retry with backoff before failing rather than treat that lag as a gate failure.
+        # A base that is a freshly created merge commit can still lag GitHub's own replication.
+        # So retry with backoff before failing, rather than treat that lag as a gate failure.
         resolved=false
         for delay in 0 2 4 8 15 30; do
           if [ "$delay" -gt 0 ]; then
@@ -109,7 +115,8 @@ runs:
           inert_excludes="${inert_excludes//$'\r'/'\x0d'}"
           echo "Excluding ${#excluded_paths[@]} path(s) from $exclude_file: ${inert_excludes//'##['/'##\x5b'}"
         fi
-        # Naming a repository with no ref, or a ref with no repository, would be worse than naming nothing, since the value exists to identify one commit.
+        # Naming a repository with no ref, or a ref with no repository, would be worse than naming nothing.
+        # That is because the value exists to identify one commit.
         # Defaulted rather than read bare, since `set -u` makes an unset variable fatal and a harness that runs this block outside Actions sets neither.
         action_repository="${ACTION_REPOSITORY:-}"
         action_ref="${ACTION_REF:-}"
@@ -129,6 +136,6 @@ runs:
           provenance_args=(--provenance "${action_repository}@${action_ref}")
         fi
         # Options first, then `--`, so every remaining token is read as a literal path.
-        # A path-shaped token beginning with a dash would otherwise parse as an option, and
-        # `--list-files` in that position turns the gate into a file listing that exits 0.
+        # A path-shaped token beginning with a dash would otherwise parse as an option.
+        # In that position, `--list-files` turns the gate into a file listing that exits 0.
         uv run --no-project --script "$GATE_PATH/prose_lint.py" --diff "$BASE" "${exclude_args[@]}" "${allow_args[@]}" "${provenance_args[@]}" -- "${scan[@]}"

.github/actions/prose-gate/prose-gate.json

@@ -10,7 +10,7 @@
     "comment-added": { "default": true, "summary": "a comment line this change adds or edits, opening its own line" },
     "dupword": { "default": true, "summary": "a duplicated consecutive word" },
     "sentence-split": { "default": false, "summary": "a sentence wrapping across lines" },
-    "sentence-length": { "default": false, "summary": "a sentence over the word cap" },
+    "sentence-length": { "default": true, "summary": "a sentence over the word cap" },
     "spelling": { "default": true, "summary": "a British spelling where the repo convention is US English" },
     "home-path": { "default": true, "summary": "an absolute home path naming a real account" },
     "dead-path": { "default": true, "summary": "a mention of a path git once tracked and the tree no longer holds" },

.github/actions/prose-gate/prose_lint.py

@@ -5,15 +5,25 @@
 # ///
 """Enforce the GOVERNANCE.md "Documentation Style Conventions" rules no other linter checks.
 
-Three layers: extract text units through Pygments and markdown-it-py, normalize them per language
-from the `prose-gate.json` beside this script, and evaluate each rule in `RULE_TABLE` on a unit. The
-JSON holds the language data, each rule's message and default, and the vocabulary and exemptions.
+Three layers:
+
+- Extract text units through Pygments and markdown-it-py.
+- Normalize them per language from the `prose-gate.json` beside this script.
+- Evaluate each rule in `RULE_TABLE` on a unit.
+
+The JSON holds the language data, each rule's message and default, and the vocabulary and
+exemptions.
 
 A repository may carry `.prose-gate.json` at its root with additions to the carried exemptions and
-nothing else, each a list merged over the carried one: `hub_hosted` and `home_path_exempt` take
-paths, `generated_trees` directory names, `service_accounts` account names, `dupword_allow` and
-`spelling_allow` phrases and words, and `tool_directives` directive names matched literally. Any
-other key is refused, since an override that silently does nothing reads as one that works.
+nothing else. Each is a list merged over the carried one:
+
+- `hub_hosted` and `home_path_exempt` take paths.
+- `generated_trees` takes directory names.
+- `service_accounts` takes account names.
+- `dupword_allow` and `spelling_allow` take phrases and words.
+- `tool_directives` takes directive names matched literally.
+
+Any other key is refused, since an override that silently does nothing reads as one that works.
 
 Exit 1 on a finding, 2 when the run cannot be scoped. Read-only. Every git read names its encoding,
 with surrogateescape, so a byte the locale cannot map neither ends the read nor changes a name.
@@ -610,8 +620,8 @@ def comment_lines(
 ) -> list[tuple[Line, bool]]:
     """Every comment line in `text`, and whether it is a documentation comment.
 
-    A comment opening on no marker the language declares belongs to a language embedded in it,
-    such as a script in HTML, and is left out. `shift` is the line `text` starts after.
+    A comment opening on no marker the language declares belongs to a language embedded in it, such
+    as a script in HTML. Such a comment is left out. `shift` is the line `text` starts after.
     """
     starts = Starts(text)
     out: list[tuple[Line, bool]] = []
@@ -707,7 +717,7 @@ def split_heredocs(text: str) -> list[tuple[int, str]]:
 
     A body's prose can hold an apostrophe, which the shell lexer reads as a string running on to the
     next one anywhere after it. The bodies come from the lexer that knows shell state, run on the
-    script without its YAML indent, and each is lexed alone so that string ends with the body.
+    script without its YAML indent. Each is lexed alone so that string ends with the body.
     """
     rows = text.splitlines(keepends=True)
     indent = min((len(r) - len(r.lstrip()) for r in rows if r.strip()), default=0)

.github/actions/pypi-build-default/action.yml

@@ -119,7 +119,8 @@ runs:
       with:
         name: pypi-build-${{ inputs.branch }}
         path: ${{ inputs.project-dir }}/dist/*
-        # A build that emits nothing would otherwise only warn, create no artifact at all, and fail one job later on a download that cannot say why.
+        # A build that emits nothing would otherwise only warn and create no artifact at all.
+        # It would fail one job later on a download that cannot say why.
         if-no-files-found: error
         # Consumed within the run by the caller stub's publish-pypi job, so minimize artifact storage.
         retention-days: 1

.github/actions/python-directories/python_directories.py

@@ -6,16 +6,20 @@
 keeps the root, where a root pyproject.toml is tracked, which is the only directory the
 validator gated before the input existed.
 
-The declaration is what decides what is gated, never discovery, since a stray or vendored
-`.py` file would otherwise change the gate silently. Tracked `.py` files that no resolved
-directory covers are therefore reported as a warning rather than gated, and the fleet audit,
-which compares the declaration against the registry, is where they become a finding.
+The declaration is what decides what is gated, never discovery, since a stray or vendored `.py` file
+would otherwise change the gate silently. Tracked `.py` files that no resolved directory covers are
+therefore reported as a warning rather than gated. The fleet audit, which compares the declaration
+against the registry, is where they become a finding.
 
-Each directory is classed by the shape that decides how its tools run: `uv` for a tracked
-uv.lock of its own, `pip` for a requirements*.txt beside it, `lint-only` for a pyproject.toml
-with no [project] or [build-system] table, and `unsupported` for a project with none of those,
-whose dependencies the validator cannot install. A uv workspace member is that last shape,
-since its lock belongs to the workspace root, which is the directory to declare instead.
+Each directory is classed by the shape that decides how its tools run:
+
+- `uv` for a tracked uv.lock of its own.
+- `pip` for a requirements*.txt beside it.
+- `lint-only` for a pyproject.toml with no [project] or [build-system] table.
+- `unsupported` for a project with none of those, whose dependencies the validator cannot install.
+
+A uv workspace member is that last shape, since its lock belongs to the workspace root, which is the
+directory to declare instead.
 
 Writes `projects` (one `<shape><TAB><directory>` line each), `any`, and `declared` to
 GITHUB_OUTPUT. Exits 1 on a declaration naming a path that is not a tracked Python project.
@@ -102,8 +106,8 @@ def in_directory(directory: str, name: str) -> str:
 def shape(directory: str, tracked_set: set[str], read: Callable[[str], str]) -> str:
     """How the directory's tools run, from its tracked files and its pyproject.toml.
 
-    The pyproject.toml is parsed first whatever the shape, so a malformed one fails here rather than
-    reading as a project with no [project] table in a later step.
+    The pyproject.toml is parsed first whatever the shape. A malformed one therefore fails here
+    rather than reading as a project with no [project] table in a later step.
     """
     config = tomllib.loads(read(in_directory(directory, "pyproject.toml")))
     if in_directory(directory, "uv.lock") in tracked_set: