How to Use Accessibility VPAT Draft from Feature Inventory (Not Legal Advice)

This guide shows how to use the PromptDig prompt Accessibility VPAT Draft from Feature Inventory (Not Legal Advice) for a real job workflow. The prompt is built so you paste locked Inputs and get a structured draft without invented facts.

What this prompt is for

Draft a VPAT-style accessibility outline from a feature inventory. Not legal advice. No invented WCAG scores or conformance claims.

Use it when you already have a source packet and need a clean first draft you can audit against that packet. It is not a substitute for counsel, clinical judgment, or an auditor.

Before you open the model

• Collect the source packet you will paste into [Source].

• Set a hard [Cap] so the draft stays usable.

• List [Banned] words and a [Never] invent list for your domain.

• Choose [Format] and keep [Lang] as English unless you intentionally change it.

• If something is unknown, put UNKNOWN in [Extra] instead of guessing.

How to run it on PromptDig

• Open the live prompt: Accessibility VPAT Draft from Feature Inventory (Not Legal Advice).

• Replace every bracketed field with your locked details.

• Run on ChatGPT, Claude, or Gemini.

• Read the honesty ledger first. Confirm Source nouns and Cap before trusting the draft body.

• Treat anything marked NOT IN INPUTS as a gap you still owe, not as model creativity.

What good output looks like

A solid run returns:

• An honesty ledger with Cap, Lang, and forbidden invent items

• A main draft that only uses nouns present in Source

• An allowlist or checklist of Source items

• A gaps list of five owed fields

• A refuse list for inventing identifiers, fees, stats, or paths

• A compliance pass that quotes Banned and Never hits

If the model invents owners, IDs, ratings, effect sizes, or paths, reject that section and tighten Never.

Tips for this job shape

• Keep Cap numeric when the job has a list or section budget.

• Paste only facts you can defend in review.

• For compliance, security, legal, clinical, or accessibility jobs, keep the not-advice banner and de-identify people.

• For image or voice compiler jobs, leave celebrity names, logos, and cloned voices out of Inputs.

Related PromptDig links

• Browse more prompts: Browse more prompts

• Share a prompt with your team: Share a prompt

• This prompt: Accessibility VPAT Draft from Feature Inventory (Not Legal Advice)

Closing checklist

Paste Source, Cap, Banned, Never, Format, Lang, and Extra. Run once. Cut Banned hits. Fill the five gaps yourself. Then ship the draft to your normal review path.

This how-to is informational only. It does not create legal, clinical, audit, or compliance advice. Always verify against your own policies and source documents.

Keep every claim auditable against Inputs. Prefer UNKNOWN over guesses. Re-run only after you add missing Source nouns. Do not ask the model to invent evidence.

Keep every claim auditable against Inputs. Prefer UNKNOWN over guesses. Re-run only after you add missing Source nouns. Do not ask the model to invent evidence.

Keep every claim auditable against Inputs. Prefer UNKNOWN over guesses. Re-run only after you add missing Source nouns. Do not ask the model to invent evidence.

Keep every claim auditable against Inputs. Prefer UNKNOWN over guesses. Re-run only after you add missing Source nouns. Do not ask the model to invent evidence.

Keep every claim auditable against Inputs. Prefer UNKNOWN over guesses. Re-run only after you add missing Source nouns. Do not ask the model to invent evidence.

Keep every claim auditable against Inputs. Prefer UNKNOWN over guesses. Re-run only after you add missing Source nouns. Do not ask the model to invent evidence.

Keep every claim auditable against Inputs. Prefer UNKNOWN over guesses. Re-run only after you add missing Source nouns. Do not ask the model to invent evidence.