Ai Alt Text Accessibility Sheet

SKILL.md

---
name: ai-alt-text-accessibility-sheet
description: "Create a batch alt text accessibility QA sheet for images before publishing, with image purpose, concise alt text, context notes, decorative-image decisions, review flags, and protected-trait safeguards."
---

# AI Alt Text Accessibility Sheet

## Purpose

Help the user turn one image or a batch of images into a visible accessibility QA artifact before publishing. The deliverable is a sheet with image label, publishing context, purpose, concise alt text, optional longer description, decorative-image decision, review flags, and publish-ready copy.

This is a prompt-only accessibility drafting workflow. It supports human review; it does not certify legal compliance, replace an accessibility audit, or guess sensitive traits from images.

## Use This Skill When

Use this skill when the user wants to:

- Draft alt text for social posts, newsletters, blogs, product pages, slides, reports, course materials, event pages, or documentation.
- Review a batch of image captions or AI-generated alt text before publishing.
- Decide whether an image needs alt text, a longer description, or should be marked decorative.
- Create a shareable spreadsheet-style QA record for editors, clients, teammates, or accessibility reviewers.
- Make image descriptions clearer, shorter, more contextual, and less biased.

Do not use this skill to identify a private person, infer protected traits, diagnose a condition, assess disability, guess age, guess ethnicity, infer religion, infer gender identity, infer health status, read hidden intent, or make claims not supported by the visible image and supplied context.

## Best Inputs

Ask only for details that improve the sheet. If some details are missing, proceed with explicit assumptions and review flags.

- Image files, screenshots, URLs, or user-provided image descriptions.
- Publishing channel: website, social media, slide deck, newsletter, app, PDF, learning material, or product page.
- Audience and purpose: inform, sell, teach, document, entertain, report, compare, or decorate.
- Surrounding text, caption, headline, product name, chart title, or link destination.
- Whether the image is essential, illustrative, decorative, redundant with nearby text, or a complex chart or diagram.
- Brand tone or style constraints: plain, warm, technical, formal, concise, playful, or multilingual handoff.
- Any known names, roles, locations, events, dates, or objects supplied by the user.
- Platform length limits or accessibility guidance the user must follow.

If images are not directly available, ask the user to upload them or provide a neutral description. Do not invent visual details.

## Workflow

1. **Set the publishing context.** Confirm where the images will appear and what nearby text already explains.
2. **Inventory the images.** Assign a stable image label such as `hero-01`, `product-02`, `chart-03`, or the filename.
3. **Identify the purpose.** Mark each image as informative, functional, complex, emotional, product-detail, decorative, redundant, or needs review.
4. **Describe only relevant visible content.** Include the details needed for the user's context. Omit visual clutter unless it matters.
5. **Avoid sensitive inference.** Do not guess identity, age, disability, health, ethnicity, religion, gender identity, pregnancy, relationship status, socioeconomic status, or other protected traits. Use neutral language such as "person," "group," or user-supplied names and roles.
6. **Draft concise alt text.** Prefer one clear sentence when possible. Put the most important information first.
7. **Handle complex images.** For charts, diagrams, maps, infographics, screenshots, tables, or dense visuals, provide short alt text plus a longer description or nearby-text recommendation.
8. **Mark decorative images.** If the image adds no information beyond surrounding text, recommend empty alt text or a decorative mark according to the user's publishing system.
9. **Add QA flags.** Note missing context, possible over-description, sensitive-trait risk, text-in-image, poor contrast, unclear purpose, or need for human subject approval.
10. **Produce the sheet.** Return a spreadsheet-style artifact that can be copied into a tracker or handed to an editor.

## Output Format

Return the artifact in this order:

1. **Batch Snapshot**

| Field | Detail |
|---|---|
| Publishing channel | |
| Audience | |
| Image count | |
| Surrounding text supplied | |
| Style or length constraints | |
| Assumptions | |

2. **Alt Text QA Sheet**

| Image label | Purpose | Draft alt text | Longer description needed? | Decorative? | Context note | QA flags |
|---|---|---|---|---|---|---|
| | | | | | | |

3. **Publish-Ready Copy**

Use this format for each image:

```text
Image label:
Alt text:
Long description, if needed:
Decorative setting, if applicable:
Editor note:
```

4. **Complex Image Notes**

For charts, diagrams, maps, screenshots, infographics, or tables, include:

- What the image is meant to communicate.
- The key data, trend, relationship, or action shown.
- What should move into surrounding text instead of being trapped in the image.
- Any missing source, label, legend, or context needed before publishing.

5. **Sensitivity and Bias Check**

Include concise checks:

- No guessed identity, health, age, disability, ethnicity, religion, gender identity, pregnancy, or protected traits.
- User-supplied names or roles are clearly marked as supplied context.
- Facial expression, body size, clothing, assistive tools, and setting are described only when relevant to the image purpose.
- Claims about products, safety, performance, or results are supported by the provided context.

6. **Open Questions**

List missing image files, publishing channel, audience, surrounding text, platform limit, product details, chart source, or reviewer decision needed before final publishing.

## Message Style

- Keep alt text concise, useful, and context-aware.
- Use neutral plain English.
- Put the user's publishing purpose ahead of generic visual description.
- Mark uncertain details as uncertain or needing review.
- Prefer "person" or "people" unless the user supplies names, roles, or relevant relationship labels.
- Avoid filler such as "image of" unless the medium matters.
- Do not include jokes, sarcasm, invented facts, hidden motives, or unsupported claims.

## Safety Boundary

- Do not identify private people or compare faces to public figures.
- Do not infer or guess sensitive traits, including age, disability, health status, ethnicity, race, religion, nationality, immigration status, gender identity, sexual orientation, pregnancy, or socioeconomic status.
- Do not diagnose visible conditions, injuries, mobility, neurodivergence, or medical needs from an image.
- Do not claim legal accessibility compliance. Present the sheet as drafting and QA support for human review.
- Do not provide instructions for scraping private images, bypassing access controls, or using images without permission.
- If an image includes minors, private people, medical scenes, identity documents, addresses, or other sensitive details, minimize unnecessary detail and flag for consent and privacy review.

## Example Prompts

- "Create an alt text QA sheet for these five blog images."
- "Review my product page alt text and make it more useful."
- "Draft alt text for this LinkedIn carousel without guessing protected traits."
- "Build a spreadsheet-style accessibility sheet for the screenshots in this tutorial."

## Install-First Success Path

**Input:** User says "Create an alt text QA sheet for these five blog images" and describes or pastes the image contexts: a chart showing quarterly revenue, a team photo at a conference, a product screenshot, an infographic about climate data, a decorative hero image.

**Steps:**
1. List each image with its context and purpose (informational, decorative, functional)
2. For each image, draft alt text that is descriptive, concise, and context-appropriate
3. Build a QA sheet with columns: Image, Context, Purpose, Draft Alt Text, WCAG Check, Notes
4. Flag any alt text that guesses protected traits (age, race, gender, disability) and suggest neutral alternatives
5. Deliver the sheet in a copy-paste-friendly format

**Output:** A structured alt text accessibility sheet. The user copies each alt text into their CMS or codebase and passes WCAG accessibility checks with confidence.