Exécutez n'importe quel Skill dans Manus
en un clic

Exécutez n'importe quel Skill dans Manus en un clic

find-style-issues

Étoiles7

Forks7

Mis à jour17 juin 2026 à 21:48

Find documentation pages that need style improvements (long paragraphs, non-sentence-case headings, over-long shell code lines, and Vale findings)

Installation

Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.

Exécuter dans Manus

Source

giantswarm

giantswarm/docs

Ouvrir le dépôt GitHub Voir les dépôts du créateur

Téléchargement

Exécuter dans Manus

Métiers associésSOC

Basé sur la classification professionnelle SOC

Analystes en assurance qualité des logiciels et testeursProfessions informatiques et mathématiques·SOC 15-1253

Explorateur de fichiers

2 fichiers

SKILL.md

readonly

name	find-style-issues
description	Find documentation pages that need style improvements (long paragraphs, non-sentence-case headings, over-long shell code lines, and Vale findings)
argument-hint	[path ...] [--vale] [--top N]

Find articles that need style improvements

Scan the documentation and rank pages by the number of style problems they contain, so you know where to point /improve-style next.

A helper script does the finding in code — don't eyeball files by hand.

What it detects

Long paragraphs — prose paragraphs whose reading length (markup, link URLs and shortcodes stripped) exceeds 600 characters.
Headings that aren't sentence case — a heuristic pre-filter. Brand names, acronyms, code identifiers and camelCase field names are allowed; Vale's Microsoft.Headings rule remains the authoritative check.
Over-long shell code lines — lines longer than 65 characters inside sh / bash / shell / console fenced code blocks.
Vale findings — error / warning / suggestion counts per file (opt-in, needs Docker).

Steps

Run the structural scan (fast, pure Python, covers all 5000+ pages):
```
python3 .claude/skills/find-style-issues/find_style_issues.py
```
By default it scans src/content and excludes auto-generated sections that aren't worth manual style review:
- src/content/changes (changelogs)
- src/content/reference/platform-api/cluster-apps
- src/content/reference/platform-api/crd
To include Vale findings, add --vale. This runs the same Docker image as the improve-style skill, so it's slower — scope it to a directory or a few files rather than the whole tree:
```
python3 .claude/skills/find-style-issues/find_style_issues.py --vale \
  src/content/tutorials/observability
```
Useful flags:
- --top N — only show the N worst files (results are ranked by Vale errors, then total issue count).
- path ... — one or more files or directories to scan instead of the default.
- --exclude SUBSTR — skip paths containing SUBSTR (repeatable).
- --no-default-excludes — also scan the generated sections listed above.
- --max-paragraph N / --max-code-line N — override the thresholds.
- --json — machine-readable output for further processing.
Report the ranked list to the user. The heading check is heuristic, so treat its hits as candidates to review, not certain defects. To actually fix a page, run /improve-style <path> on it.

Example

# Worst 15 pages in the tutorials section, including Vale
python3 .claude/skills/find-style-issues/find_style_issues.py --vale --top 15 \
  src/content/tutorials

Output is grouped per file, e.g.:

src/content/overview/security/platform-security/index.md
  long-paragraph : 2 (longest 1137 chars) — L55 (762), L180 (1137)
  heading-case   : L53 — word 'Operator' is capitalized mid-heading: "Trivy Operator"
  long-code-line : 16 (longest 158 chars) — L81 (158), L82 (158), ...
  vale           : 23 (6 error, 0 warning, 17 suggestion)

Plus depuis ce dépôt

même dépôt

publish-crd-reference

giantswarm/docs

Understand and maintain how the CRD reference pages under /reference/platform-api/crd/ are generated — register a new CRD source repo, make sure every published CRD has an example CR, and fix pages that render without an example

2026-07-037

classify-diataxis

giantswarm/docs

Classify documentation pages into their Diátaxis type and propose the diataxis_content_type frontmatter value, flagging pages that mix modes

2026-06-267

export-to-miro

giantswarm/docs

Create a box on a Miro board for every row in a CSV inventory of pages (e.g. the output of the export-csv skill) — each box shows the page title as a clickable link to its URL, with the navigation path as a subline

2026-06-267

export-csv

giantswarm/docs

Export a CSV inventory of all documentation pages (title, navigation path levels, and public URL) from the frontmatter of every Markdown file under src/content

2026-06-267

assess-readability

giantswarm/docs

Assess English readability of documentation pages (Flesch Reading Ease, Flesch-Kincaid grade, Gunning Fog, LIX), excluding generated content

2026-06-227

improve-style

giantswarm/docs

Apply style fixes and run Vale prose linting on modified or added pages

2026-06-227

Find articles that need style improvements

Scan the documentation and rank pages by the number of style problems they contain, so you know where to point /improve-style next.

A helper script does the finding in code — don't eyeball files by hand.

What it detects

Long paragraphs — prose paragraphs whose reading length (markup, link URLs and shortcodes stripped) exceeds 600 characters.

Headings that aren't sentence case — a heuristic pre-filter. Brand names, acronyms, code identifiers and camelCase field names are allowed; Vale's Microsoft.Headings rule remains the authoritative check.

Over-long shell code lines — lines longer than 65 characters inside sh / bash / shell / console fenced code blocks.

Vale findings — error / warning / suggestion counts per file (opt-in, needs Docker).

Steps

Run the structural scan (fast, pure Python, covers all 5000+ pages):

python3 .claude/skills/find-style-issues/find_style_issues.py

By default it scans src/content and excludes auto-generated sections that aren't worth manual style review:

src/content/changes (changelogs)
src/content/reference/platform-api/cluster-apps
src/content/reference/platform-api/crd

To include Vale findings, add --vale. This runs the same Docker image as the improve-style skill, so it's slower — scope it to a directory or a few files rather than the whole tree:

python3 .claude/skills/find-style-issues/find_style_issues.py --vale \
  src/content/tutorials/observability

Useful flags:

--top N — only show the N worst files (results are ranked by Vale errors, then total issue count).
path ... — one or more files or directories to scan instead of the default.
--exclude SUBSTR — skip paths containing SUBSTR (repeatable).
--no-default-excludes — also scan the generated sections listed above.
--max-paragraph N / --max-code-line N — override the thresholds.
--json — machine-readable output for further processing.

Report the ranked list to the user. The heading check is heuristic, so treat its hits as candidates to review, not certain defects. To actually fix a page, run /improve-style <path> on it.

Example

# Worst 15 pages in the tutorials section, including Vale python3 .claude/skills/find-style-issues/find_style_issues.py --vale --top 15 \ src/content/tutorials

Output is grouped per file, e.g.:

src/content/overview/security/platform-security/index.md long-paragraph : 2 (longest 1137 chars) — L55 (762), L180 (1137) heading-case : L53 — word 'Operator' is capitalized mid-heading: "Trivy Operator" long-code-line : 16 (longest 158 chars) — L81 (158), L82 (158), ... vale : 23 (6 error, 0 warning, 17 suggestion)