Ai Source Claim Audit Tags

Create printable or digital audit tags for reviewing AI-assisted drafts at claim level, including source-needed flags, unsupported-number stickers, quote-check tabs, final-approval lane cards, and a human verification checklist without browsing, fabricating sources, or verifying facts automatically.

Overview

Use this skill when the user has an AI-generated or AI-assisted memo, essay, article, report, proposal, client brief, slide outline, draft email, or school draft that contains claims requiring human source review before sharing or submission.

This is a prompt-only document organization skill. It creates printable or digital review artifacts that help the user mark claims, numbers, quotes, names, dates, recommendations, assumptions, and unsupported statements. It does not verify facts, browse the web, create hidden citations, fabricate sources, submit work, or bypass academic or workplace rules.

Trigger

Use this skill when the user asks for:

• AI claim audit tags, source checklist, citation review labels, or fact-check worksheet.
• Stickers or tabs for marking source-needed claims in an AI draft.
• A review system for AI-generated statistics, quotes, names, dates, recommendations, or claims.
• A final source approval checklist before sending, publishing, submitting, or presenting AI-assisted text.

Do not use this skill to browse for sources, invent citations, create fake references, guarantee accuracy, submit work, or help the user evade disclosure, citation, academic, workplace, or publication rules.

Intake

Ask only what is needed to build the review artifact:

• What kind of document is being reviewed, such as memo, essay, article, report, proposal, client brief, slide outline, draft email, or school draft.
• How the user will review it, such as printed pages, shared document, notes app, slide deck, draft email, or pasted AI output.
• Which claim types matter most, such as statistics, dates, names, prices, rankings, policy references, quotes, expert claims, recommendations, comparisons, or legal or medical adjacent statements.
• Whether the user wants printable labels, digital tags, a checklist, lane cards, or a compact combined sheet.
• Any known review standard, such as school citation rules, workplace source policy, editor requirements, or client approval process.

If the user provides draft text, use it only to create review labels and a claim review plan. Do not state that any claim is true unless the user has already provided trusted verification.

Workflow

1. Confirm the review scope and remind the user that all claims need human verification with trusted sources before external use.
2. Identify the review surface, such as printed pages, shared document comments, slide notes, email draft, notes app, or pasted AI output.
3. Create audit tags for sourced, source needed, verify date, verify number, check name, quote needed, opinion, assumption, and remove.
4. Generate source-needed flags with fields for claim text, source type needed, reviewer, verification status, and decision.
5. Build unsupported-number stickers for percentages, rankings, prices, dates, totals, measurements, forecasts, and comparisons.
6. Add quote-check tabs for direct quotes, paraphrases, attributed statements, expert claims, policy references, and rule references.
7. Create final-approval lane cards for verified, revise, remove, needs disclaimer, cite properly, and ask owner.
8. Provide a one-page AI source review checklist that separates human-verified claims from text that still needs revision, removal, citation, or disclaimer wording.

Output Format

Return these sections:

1. Review Snapshot: document type, review surface, review goal, claim types to prioritize, and scope limits.
2. Audit Tag Sheet: concise tag labels with short instructions for where each tag belongs.
3. Source-Needed Flags: reusable flag templates with fields for claim text, source type needed, reviewer, verification status, and decision.
4. Unsupported-Number Stickers: labels for numbers, dates, rankings, prices, totals, measurements, forecasts, and comparisons.
5. Quote-Check Tabs: tabs for direct quote, paraphrase, attribution, expert claim, policy reference, and rule reference.
6. Final-Approval Lane Cards: verified, revise, remove, needs disclaimer, cite properly, and ask owner.
7. AI Source Review Checklist: manual steps before sharing, submitting, publishing, sending, or presenting the draft.
8. Boundary Notes: no automatic fact verification, no fabricated citations, no hidden sources, and follow applicable school, workplace, client, or publication rules.

For a quick request, provide a compact tag sheet plus the final checklist first.

Tag Guidance

Use clear, short labels that fit on printed tabs, sticky notes, document comments, or spreadsheet cells. Suggested labels:

• SOURCED: A trusted source has been checked and recorded.
• SOURCE NEEDED: The claim needs a source before external use.
• VERIFY DATE: The date, deadline, period, or timeline needs checking.
• VERIFY NUMBER: The statistic, price, percentage, count, or measurement needs checking.
• CHECK NAME: A person, organization, place, product, title, or spelling needs checking.
• QUOTE NEEDED: Exact wording or citation is required before using as a quote.
• OPINION: This is judgment, interpretation, or recommendation, not a sourced fact.
• ASSUMPTION: This depends on unstated context or unverified inference.
• REMOVE: This should leave the draft unless verified or rewritten.

Review Rules

• Work only from user-provided text, user-stated context, or the requested label system.
• Do not browse, fetch sources, or imply that any claim has been verified automatically.
• Do not create, complete, or format fake citations.
• Do not invent source titles, authors, URLs, page numbers, dates, quotations, or institutions.
• Mark unknowns clearly as needs verification.
• Separate factual claims, opinions, assumptions, recommendations, quotes, and paraphrases.
• Treat numbers, names, dates, rankings, prices, totals, comparisons, legal or medical adjacent statements, and policy references as high-priority review items.

Boundary Rules

• Do not submit work, send messages, post content, or approve the draft for external use.
• Do not bypass citation rules, disclosure rules, academic integrity policies, workplace policies, client requirements, or publication standards.
• Do not provide legal, medical, financial, or academic-integrity advice as a final determination.
• If the draft will be used in a high-stakes context, tell the user to verify with qualified sources or appropriate reviewers before relying on it.
• If the user asks for hidden citations or fabricated sources, refuse and offer a transparent source review checklist instead.

Acceptance Criteria

1. Produces printable or digital audit tags for sentence-level review of AI-assisted drafts.
2. Includes source-needed flags, unsupported-number stickers, quote-check tabs, final-approval lane cards, and a one-page checklist.
3. Separates verified, unverified, opinion, assumption, quote, paraphrase, and remove decisions.
4. Requires human verification with trusted sources before external use.
5. Does not browse, verify facts automatically, fabricate sources, submit work, or bypass rules.
6. Requires no code execution, APIs, network access, credentials, packages, or extra files.

Example Prompts

• "Make AI source checklist tags for this report draft."
• "I need source-needed stickers for an AI-generated essay."
• "Create a fact-check worksheet for AI claims in my slide outline."
• "Give me printable tags for statistics, quotes, names, and dates in an AI memo."
• "Build a final source approval checklist before I send this AI-assisted brief."