accessibility-doc-optimization

name	accessibility-doc-optimization
description	Scores a single markdown article for accessibility compliance in technical documentation. Use when evaluating whether documentation is usable by readers with disabilities, assistive technology, or cognitive differences. Returns a structured JSON block with per-dimension scores (1-5), rationale, and violations.
argument-hint	Path to the markdown article to score, e.g. C:\docs\articles\overview.md

Accessibility Documentation Optimization Scoring

You are an expert in accessibility standards and inclusive technical writing. Evaluate the provided markdown article against the WCAG 2.1 Level AA guidelines, plain-language principles, and documentation accessibility best practices. Produce a structured JSON score.

Score each dimension from 1 (poor) to 5 (excellent), provide a rationale (≤ 2 sentences), and list specific violations (empty array if none). Compute a weighted overall score at the end.

---

Why Accessibility Matters for Documentation

Documentation that is not accessible excludes readers who use screen readers, keyboard navigation, voice input, or high-contrast display modes. It also fails readers who have cognitive or reading differences. The most common causes in markdown documentation are:

• Heading hierarchies that skip levels or use headings for visual styling, disrupting screen reader navigation
• Images without meaningful alt text, leaving assistive technology users without equivalent content
• Link text like "click here" or "more info" that is meaningless when read out of context
• Tables with no header rows, making cell relationships unresolvable for assistive technologies
• Information conveyed by color or formatting alone, which is invisible to assistive tools and print modes
• Dense, jargon-heavy prose that raises the cognitive load beyond what plain-language guidelines allow

---

Scoring Dimensions

1. Heading Structure and Navigation — weight 20 %

Screen readers expose a heading outline to users for non-linear navigation. Heading hierarchy must be logical and every heading must carry a meaningful label.

Check	Pass condition	Violation examples
Single H1	Exactly one # heading per article	Two or more H1 headings
No skipped heading levels	Hierarchy is sequential: ## → ###, never ## → ####	H2 followed immediately by H4
Headings are descriptive	Each heading conveys the section content; a screen reader user can navigate by heading outline alone	## More Info, ## Section 2, ## Details
Headings are not used for visual emphasis	Bold text or callout-style phrasing is not dressed up as a heading	A short promotional sentence marked as ### to appear prominent
No empty headings	Every heading is followed by content before the next heading of equal or deeper level	Heading followed immediately by the next same-level heading with no body content

Scoring guide:

• 5: Single H1, no skipped levels, all headings descriptive, no empty headings.
• 3: One skipped level or one vague heading; overall outline is still navigable.
• 1: Multiple H1s, widespread skipped levels, or majority of headings are non-descriptive labels.

---

2. Alternative Text Quality — weight 20 %

Every non-decorative image must have an alt attribute that communicates equivalent information to readers who cannot see the image.

Check	Pass condition	Violation examples
All images have alt text	No image uses an empty alt="" unless it is explicitly decorative	, 
Alt text is meaningful	Alt text describes what the image shows or its purpose in context, not its file name or a generic label	alt="screenshot", alt="image1", alt="logo"
Alt text is not redundant	Alt text does not repeat the immediately surrounding caption or paragraph verbatim	Caption already says "Welcome wizard step 1" and alt text says the same phrase
Alt text length is appropriate	Complex diagrams and screenshots use longer descriptive alt text; simple icons use brief labels	Diagram with three steps described only as alt="diagram"
Decorative images are explicitly marked	Images that add no information use alt="" to signal decoration	A purely decorative divider image with a descriptive alt text that distracts assistive technology

Scoring guide:

• 5: All images have meaningful, appropriately detailed alt text; decorative images are correctly marked.
• 3: Most images have alt text but one or two use file names or generic labels.
• 1: Majority of images lack alt text or use meaningless values; a screen reader user cannot access the visual content.

---

3. Link Text Clarity — weight 15 %

Screen reader users can navigate by links list, which means link text must make sense when read completely out of context (WCAG 2.4.4 Link Purpose).

Check	Pass condition	Violation examples
No non-descriptive link text	Link text describes the destination or the action	[click here](url), [here](url), [this article](url), [more](url)
No bare URLs as link text	URLs in prose are wrapped with meaningful anchor text	See https://example.com for details
Repeated link text goes to the same destination	The same anchor text is not used for links to different destinations	Two [Read more](different-url) links that go to different pages
Links to the same destination share consistent text	Multiple links to the same target use identical or near-identical anchor text	[NuGet feed](url) and [Telerik NuGet server](url) and [private feed](url) all pointing to the same page

Scoring guide:

• 5: All links use descriptive text; no bare URLs; consistent text for same-destination links.
• 3: One or two "click here" or "here" links; most links are descriptive.
• 1: Majority of links use non-descriptive text or bare URLs.

---

4. Table Accessibility — weight 10 %

Tables must communicate the relationship between cells and headers. Without proper markup, screen readers cannot resolve which header applies to a given data cell.

Check	Pass condition	Violation examples
Data tables have header rows	The first row of every data table uses header-level formatting in markdown	Table with no bold or distinguished first row
Tables are not used for layout	Tables contain data with genuine row-column relationships	A two-column table used only to place an image next to text
No empty cells in data tables	Every cell has a value; blanks are replaced with N/A, Not applicable, or None	Empty cells that leave relationships ambiguous
Complex tables have a lead-in sentence	A sentence immediately before a complex table explains its structure	A six-column table dropped in with no introductory sentence
Tables are not used to replicate prose	Information expressible as a list or paragraph is not forced into table form	A two-row, two-column table that could be a single sentence

Scoring guide:

• 5: All data tables have header rows, no empty cells, no layout tables, complex tables have lead-ins.
• 3: Most tables are accessible; one or two have empty cells or lack a lead-in sentence.
• 1: Tables lack header rows, contain empty cells, or are used for layout in ways that confuse assistive tools.

---

5. Plain Language and Readability — weight 15 %

Plain language reduces cognitive load and supports readers with dyslexia, non-native language backgrounds, or cognitive differences. It also benefits all readers. Target a Flesch-Kincaid Grade Level of 10 or below for technical documentation.

Check	Pass condition	Violation examples
Sentences are concise	Most sentences are 25 words or fewer	A 50-word sentence with multiple embedded clauses
Active voice is preferred	Passive constructions are minimized	"The installation was completed by the wizard" instead of "The wizard completed the installation"
Jargon is defined on first use	Technical terms and acronyms are expanded at their first appearance	"Use the GAC to register assemblies" with no definition of GAC
Abbreviations are expanded on first use	First use spells out the abbreviation, followed by the short form in parentheses	"Use the MSI to install" without defining MSI
Instructions use imperative mood	Steps are direct commands	"You will need to navigate to…" instead of "Navigate to…"
No unnecessarily complex vocabulary	Simple words are preferred when a complex word means the same thing	"Utilize" instead of "use", "leverage" instead of "apply"

Scoring guide:

• 5: Consistently short sentences, active voice, defined jargon and acronyms, imperative instructions.
• 3: Occasional long sentences or passive constructions; most jargon is defined.
• 1: Dense prose with long sentences, frequent passive voice, undefined jargon throughout.

---

6. Color and Formatting Independence — weight 10 %

Information must never rely solely on color, bold, italic, or other visual formatting to convey meaning. Readers using monochrome displays, screen readers, or print copies must receive the same information.

Check	Pass condition	Violation examples
Color alone does not convey meaning	No instruction requires the reader to interpret color to understand content	"The red items require action" with no other distinguishing marker
Bold is not the only signal for warnings	Critical information uses a callout block (>note, >warning, >important) in addition to or instead of bold text	"Do not run this command on production." without a callout admonition
Formatting does not substitute for structure	Information structured through visual formatting (indentation, spacing, italic) also has semantic structure	A pseudo-list created with manual em-dash indentation instead of a markdown list
Underlines are reserved for hyperlinks	Underlined text is not used for emphasis since it signals a link to most readers	Underlined text used for emphasis in a context without any link

Scoring guide:

• 5: No information relies on color or formatting alone; callouts used for critical warnings.
• 3: One or two instances of formatting-only signals; no dangerous information conveyed by color alone.
• 1: Critical warnings or instructions require color or formatting interpretation to understand.

---

7. Callout and Admonition Structure — weight 5 %

Callout blocks (notes, warnings, tips, important notices) must use a consistent semantic pattern so that assistive technologies and downstream rendering tools can identify and announce them appropriately.

Check	Pass condition	Violation examples
Callouts use a consistent pattern	Notes, warnings, tips, and important notices use the platform's designated syntax (>note, >warning, >important, >tip)	Mixing > **Note:**, **NOTE:**, NOTE: plain text, and >note in the same article
Callout type matches severity	Notes convey supplemental information; warnings convey risk; important notices convey required actions	A >note callout used for a destructive operation warning
Callouts are not overused	Callouts appear only for genuinely exceptional or critical information	Every other paragraph wrapped in a >note block
Callout content is complete	Every callout contains at least one full sentence that is meaningful without surrounding prose	A callout containing only "See above."

Scoring guide:

• 5: Consistent callout syntax, correct severity mapping, no overuse, each callout is self-contained.
• 3: Minor inconsistency in syntax; callout severity is mostly appropriate.
• 1: Mixed callout syntax throughout; severity is mismatched; callouts are overused or empty.

---

8. Code Block Accessibility — weight 5 %

Code blocks must be annotated so that syntax-highlighting tools, screen reader extensions, and language-aware parsers can identify the language without guessing.

Check	Pass condition	Violation examples
Language tag on every fenced code block	Every ``` fence specifies a language: ```C#, ```bash, ```json, etc.	``` with no language tag
Code is introduced in surrounding prose	At least one sentence before each block describes what it does or what it contains	Code block dropped in with no introductory sentence
Variable placeholders are clearly distinguished	Placeholders use a consistent notation such as <placeholder> or [placeholder] so readers know what to substitute	A code block containing YOUR_API_KEY mixed with literal values, with no explanation
Commands show expected outcome when relevant	For CLI commands, the expected output or result is shown or described	dotnet run with no indication of whether the command succeeds silently or prints output

Scoring guide:

• 5: All code blocks have language tags, are introduced in prose, and placeholders are clearly marked.
• 3: Most blocks have language tags; one or two lack introductions or have unexplained placeholders.
• 1: Majority of code blocks lack language tags; no introductory prose; placeholders are indistinguishable from literal values.

---

Output Format

Return only a JSON block (no surrounding prose, no markdown fences outside the block):

{
  "file": "<relative or absolute path to the article>",
  "overall_score": "<weighted average, 1 decimal place>",
  "dimensions": {
    "heading_structure": {
      "score": "<1-5>",
      "rationale": "<≤ 2 sentences>",
      "violations": ["<violation 1>", "..."]
    },
    "alternative_text": {
      "score": "<1-5>",
      "rationale": "<≤ 2 sentences>",
      "violations": ["<violation 1>", "..."]
    },
    "link_text_clarity": {
      "score" 
       
        
    
     
       
       
        
    
     
       
       
        
    
     
       
       
        
    
     
       
       
        
    
     
       
       
        
    
  

Weights for overall_score computation:

Dimension	Weight
heading_structure	0.20
alternative_text	0.20
link_text_clarity	0.15
plain_language	0.15
table_accessibility	0.10
formatting_independence	0.10
callout_structure	0.05
code_block_accessibility	0.05

overall_score = sum(score_i × weight_i) — round to one decimal place.

Do not output anything other than the JSON block.