docs-media — Media Analysis

Workflow#

1. Gather the docs — get the list of pages in scope and read their content. If a semantic/graph search tool over the markdown is available (self-hosted markdown-lsp, or a connected Docsbook workspace), prefer it — faster and cheaper than scanning files; otherwise read the files directly with grep/find. Prioritize Tier 1 pages (quick-start, pricing, auth, install) first.
2. Check scope — if no image or video references are found in the page content, skip this skill.
3. Apply checklist — from the page content / graph: detect missing alt text, generic filenames, missing captions. For physical file checks (size, age, dimensions), note that local repo access is required.
4. Produce report — return one JSON issue object per finding, sorted by severity. Note which checks required local access and were skipped.

Guardrails#

• Do not edit any documentation files — surface findings only.
• Physical file checks (size in KB, file age, image dimensions) require local repo access — note when these are skipped.
• Skip this skill entirely if no media references are found in the doc content.
• Empty alt (![]()) is correct for decorative images — flag only when surrounding context implies the image is informative.
• Converting PNG diagrams to Mermaid is a recommendation, not a requirement — ask the user before flagging as an issue.

Inputs#

This skill needs two things, by whatever means are available:

• The list of pages in scope — a docs folder, a sitemap, or a doc graph.
• The content of each page — read on demand.

Acceleration (optional). Graph/semantic search over the docs makes navigation faster and cheaper than scanning files. You can self-host it with markdown-lsp, or get the same capability in the cloud by connecting a Docsbook workspace. With nothing connected, plain file reads and grep/find work fine.

Checklist#

File Formats#

• PNG for UI screenshots — not JPG (JPG degrades text rendering)
• GIF only for short animations — ≤ 5 seconds, ≤ 2 MB
• No video files committed to the repo — use YouTube, Loom, or Vimeo embeds
• Mermaid preferred over static PNG for diagrams and flowcharts

File Size (local repo check)#

• PNG screenshot < 500 KB
• GIF < 2 MB
• Total media per page < 5 MB
• No MP4/MOV files committed to repo

# Find large media files
find . -type f \( -name '*.png' -o -name '*.jpg' -o -name '*.gif' -o -name '*.mp4' \) -size +500k -exec ls -lh {} \;

File Names (detectable from markdown links)#

• kebab-case — workspace-settings-api-keys.png
• Descriptive — name describes content, not position: not screenshot1.png, image.png, Screenshot 2024-01-15.png
• No spaces or special characters in file names
• No timestamps in file name — content description, not capture date
• Prefix by type optionally — ui-, diagram-, icon-

Alt Text (detectable from markdown)#

• All informative images have alt — ![alt](image.png) not ![](image.png)
• Decorative images have empty alt — ![](decoration.png)
• Alt describes content — "Workspace settings with API key field" not "Screenshot"
• Alt is not "image of" / "screenshot of" — screen reader already announces it
• Length ≤ 125 characters — longer descriptions go in the body

UI Screenshots#

• Cropped to the relevant area — not the full browser window with chrome
• No personal data — use test accounts, placeholder names/emails
• No timestamps visible on the screenshot itself (they go stale)
• Important area is highlighted — box, arrow, or highlight overlay
• Consistent theme — all screenshots in light mode OR all in dark mode

Screenshot Freshness (local repo check)#

Screenshots older than 180 days in an actively developed product are very likely stale:

# Screenshots older than 180 days
find . -name '*.png' -mtime +180 -exec ls -lh {} \;

From the page content / graph: check last_updated of pages containing screenshots (from graph metadata if available) — if a page hasn't been updated in 180+ days and contains screenshots, flag for manual review.

Videos#

• Captions/subtitles on all videos
• Transcript or summary in the page text
• Not auto-play
• Described in text before the embed — what the video shows, how long it is
• Hosted externally — YouTube, Loom, Vimeo — not committed as a file

Diagrams#

• Mermaid in markdown for flowcharts, architecture, ER diagrams:

  ```mermaid
  graph LR
    GitHub --> Docsbook
    Docsbook --> CDN
    CDN --> Users
  ```
  

• Source file saved alongside PNG export — .excalidraw, .drawio — so diagrams can be updated
• Text alternative for complex diagrams that can't be Mermaid
• Readable in both light and dark mode

What to Look For#

Output Format#

{
  "file": "docs/quick-start.md",
  "line": 34,
  "severity": "critical",
  "rule": "image-missing-alt",
  "found": "Line 34: ![](screenshots/setup.png) — informative image with no alt text.",
  "suggestion": "Add descriptive alt: '![Docsbook workspace creation form with GitHub repo URL field](screenshots/setup.png)'"
}

{
  "file": "docs/guides/api-setup.md",
  "line": 67,
  "severity": "high",
  "rule": "generic-filename",
  "found": "Line 67: ![Settings panel](screenshots/screenshot1.png) — filename 'screenshot1.png' is not descriptive and will be impossible to identify when the folder has many screenshots.",
  "suggestion": "Rename to a descriptive kebab-case name: 'screenshots/api-settings-key-field.png'. Update the link in the markdown."
}

{
  "file": "docs/architecture.md",
  "line": 12,
  "severity": "medium",
  "rule": "diagram-could-be-mermaid",
  "found": "Line 12: ![System architecture](architecture.png) — static PNG diagram. Hard to keep updated without design tools.",
  "suggestion": "Convert to Mermaid:\n```mermaid\ngraph LR\n  GitHub --> Docsbook\n  Docsbook --> CDN\n  CDN --> Users\n```\nMermaid renders automatically, versions in git, and updates in a PR without design software."
}

Acceptance Criteria#

• Every image reference in scope has been checked for alt text and filename quality.
• Physical file size and age checks are either run (local access) or explicitly noted as skipped with a reason.
• Skill exits early with a clear message if no media references are found.
• Output is valid JSON per the format above, one object per finding.

Related Skills#

• docs-accessibility — alt text is an a11y requirement
• docs-maintenance — stale screenshots are a maintenance issue
• docs-seo — descriptive filenames and alt text help SEO
• docs-analyze — orchestrator