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

2. Alt Text QA Sheet

3. Publish-Ready Copy

Use this format for each image:

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.