AGENCYSCRIPT
CoursesEnterpriseBlog
đź‘‘FoundersSign inJoin Waitlist
AGENCYSCRIPT

Governed Certification Framework

The operating system for AI-enabled agency building. Certify judgment under constraint. Standards over scale. Governance over shortcuts.

Stay informed

Governance updates, certification insights, and industry standards.

Products

  • Platform
  • Certification
  • Launch Program
  • Vault
  • The Book

Certification

  • Foundation (AS-F)
  • Operator (AS-O)
  • Architect (AS-A)
  • Principal (AS-P)

Resources

  • Blog
  • Verify Credential
  • Enterprise
  • Partners
  • Pricing

Company

  • About
  • Contact
  • Careers
  • Press
© 2026 Agency Script, Inc.·
Privacy PolicyTerms of ServiceCertification AgreementSecurity

Standards over scale. Judgment over volume. Governance over shortcuts.

On This Page

Phase One: Setup Before You PromptName the ReaderState the PurposeIdentify the Failure You Most Want to AvoidPhase Two: The Prompt ItselfSet a Concrete LengthChoose a Structure That Matches ConsumptionSeparate Instructions From SourcePhase Three: Fidelity ConstraintsList What Must Survive CompressionForbid InventionForbid Confidence InflationPhase Four: Verification After OutputRead for Omission Against the SourceSpot-Check Specifics and TonePromote Recurring Fixes Into the TemplateTailoring the Checklist by Document TypeContracts and AgreementsTranscripts and Meeting NotesResearch and Analytical DocumentsHow to Use the Checklist Without It Becoming a ChoreTurning the Checklist Into a Team HabitMake the Checklist Visible at the Point of SendingReview What the Checklist Keeps CatchingKeep It Short Enough to Actually RunFrequently Asked QuestionsDo I really need to run all twelve items every time?Which phase catches the most problems?How is this different from just writing a good prompt?Can I adapt the checklist for specific document types?Key Takeaways
Home/Blog/Run This Before You Trust Any AI Summary in 2026
General

Run This Before You Trust Any AI Summary in 2026

A

Agency Script Editorial

Editorial Team

·February 12, 2022·6 min read
prompting for summarization qualityprompting for summarization quality checklistprompting for summarization quality guideprompt engineering

A checklist is only useful if you understand why each item is on it, otherwise you skip the ones that feel optional and they are usually the ones that matter. This checklist is built for actual use: every item comes with a one-line justification so you know what you are protecting against when you check it off. Treat it as a pre-flight pass you run before sending any summary that carries consequences.

The items are grouped into four phases that follow the natural order of work: setup before you prompt, the prompt itself, the constraints that guard fidelity, and verification after the output appears. You can run the whole thing in a couple of minutes once it is familiar, and you can safely shorten it for low-stakes summaries.

Copy it, adapt it to your document types, and keep it somewhere you will actually see it when you are about to hit send.

Phase One: Setup Before You Prompt

These items take seconds and prevent the largest category of failures.

Name the Reader

Confirm you have stated who the summary is for. Without an audience, the model produces a generic middle that serves no one and usually needs rewriting.

State the Purpose

Confirm you have said what the reader will do with the summary. Purpose drives length, tone, and selection, so naming it lets the model make the right downstream choices automatically.

Identify the Failure You Most Want to Avoid

Decide in advance what would make this summary unacceptable, a dropped deadline, an inflated conclusion, a buried risk. That choice tells you which constraint to emphasize.

Phase Two: The Prompt Itself

These items shape what the model produces.

Set a Concrete Length

Confirm the length is a number, not a soft word. Enforceable targets like "under 120 words" produce consistent output; "keep it short" does not.

Choose a Structure That Matches Consumption

Confirm you picked list or prose based on how the reader will use it. Scanning for actions wants a list; understanding a situation wants prose. The wrong structure makes the reader do the work the summary should save.

Separate Instructions From Source

Confirm a clear divider sits between your directions and the source text. Without it, the model may summarize your instructions or ignore your constraints.

Phase Three: Fidelity Constraints

These items are where summary quality is won or lost.

List What Must Survive Compression

Confirm you named the specifics to preserve, figures, dates, names, commitments. Models drop specifics by default; only an explicit list protects them.

Forbid Invention

Confirm the prompt says to use only information in the source and to flag ambiguity rather than resolve it. This prevents confident fabrication, the most damaging failure.

Forbid Confidence Inflation

Confirm the prompt requires matching the source's level of certainty. This stops the model from rewriting a hedged, preliminary finding into a settled conclusion.

Phase Four: Verification After Output

These items catch what the prompt missed.

Read for Omission Against the Source

Skim the original, list the points you would have kept, and confirm each appears. Most quality failures are omissions, which are invisible in the output and only show up in this comparison.

Spot-Check Specifics and Tone

Confirm numbers match exactly and no claim is stronger than the source. A single wrong figure or inflated conclusion can do more damage than a clumsy sentence.

Promote Recurring Fixes Into the Template

If you found yourself adding the same correction, fold it into your saved prompt. A template that absorbs corrections keeps getting better instead of repeating the same miss.

Tailoring the Checklist by Document Type

The generic checklist is a starting point. Its real power shows when you adapt the inclusions and verification items to the documents you actually handle.

Contracts and Agreements

Add an inclusion item for obligations, deadlines, penalties, and termination conditions, and a verification item that confirms exact wording survived where it carries legal weight. For these documents, a paraphrase that shifts a single condition can change the meaning, so verbatim preservation moves from optional to mandatory.

Transcripts and Meeting Notes

Add an inclusion item for every request, decision, and open question, and a verification item that scans specifically for side concerns the model tends to drop. Transcripts bury important asides in conversational noise, so the checklist has to actively hunt for them.

Research and Analytical Documents

Add an inclusion item for caveats, sample sizes, and limitations, and a verification item that compares the certainty of the summary against the source. Analytical documents are where confidence inflation does the most harm, so the checklist should treat preserved uncertainty as a pass condition.

How to Use the Checklist Without It Becoming a Chore

Run the full list only on consequential summaries, anything client-facing or decision-driving. For personal notes, the setup and a quick omission check are enough. The phases are ordered so that the cheapest, highest-value items come first, which means even a partial run captures most of the benefit. Over a few weeks the early phases become automatic, and the checklist mostly serves as a reminder for the verification steps people are most tempted to skip.

Turning the Checklist Into a Team Habit

A checklist that lives in one person's head helps one person. A little structure turns it into something a whole team relies on.

Make the Checklist Visible at the Point of Sending

The checklist works best when it sits where the decision happens, pinned near the tool people use to summarize or pasted into the template itself. Out of sight, it gets skipped under deadline pressure. In sight, it becomes a quick reflex run before hitting send.

Review What the Checklist Keeps Catching

Once a month, look at which items are still catching problems. If the verification read keeps surfacing dropped commitments, that is a signal to strengthen the inclusions line in your prompts so the failure stops happening upstream. The checklist is not only a gate; it is a feedback loop that tells you where your prompts are weak.

Keep It Short Enough to Actually Run

A checklist that grows to thirty items will be abandoned. Hold it to the dozen or so that earn their place, and prune any item that has not caught a real problem in months. Usefulness, not completeness, is what keeps a checklist alive.

Frequently Asked Questions

Do I really need to run all twelve items every time?

No. Run the full set for summaries that carry consequences and a shortened version for low-stakes work. The phases are ordered by value, so even a partial pass catches the most important issues. The point is deliberate skipping, not skipping by default.

Which phase catches the most problems?

The fidelity constraints and the omission check together catch the majority. Most summary failures are either dropped specifics or smoothed-over uncertainty, and those two areas address both. If you only have time for part of the list, run those.

How is this different from just writing a good prompt?

The prompt handles phases one through three. The checklist adds the verification phase that prompts cannot enforce, plus a reminder to promote fixes back into your template. It turns a one-time good prompt into a repeatable, improving practice.

Can I adapt the checklist for specific document types?

Yes, and you should. Add document-specific preservation items, obligations for contracts, open questions for transcripts, limitations for research. A tailored checklist per type catches the omissions that generic checks miss.

Key Takeaways

  • Every checklist item carries a justification, so you know what each one protects against.
  • Setup items, naming reader, purpose, and the failure to avoid, prevent the largest class of problems cheaply.
  • Fidelity constraints protect specifics, forbid invention, and block confidence inflation.
  • The verification phase catches omissions that prompts alone cannot enforce.
  • Run the full list for consequential summaries and a trimmed version for low-stakes ones.

Build the prompt this checklist verifies with A Step-by-Step Approach to Prompting for Summarization Quality, understand the reasoning behind each item in Prompting for Summarization Quality: Best Practices That Actually Work, and see the failure modes it guards against in 7 Common Mistakes with Prompting for Summarization Quality (and How to Avoid Them).

Search Articles

Categories

OperationsSalesDeliveryGovernance

Popular Tags

prompt engineeringai fundamentalsai toolsthe difference between AIMLagency operationsagency growthenterprise sales

Share Article

A

Agency Script Editorial

Editorial Team

The Agency Script editorial team delivers operational insights on AI delivery, certification, and governance for modern agency operators.

Related Articles

General

Prompt Quality Decides Whether AI Earns Its Keep

Prompt quality is the single biggest variable in whether AI delivers real work or expensive noise. The model matters, the platform matters — but the prompt you write determines whether you get a first

A
Agency Script Editorial
June 1, 2026·10 min read
General

Counting the Real Cost of Every Token You Send

Tokens and context windows sit at the intersection of AI capability and operational cost—yet most business cases treat them as technical footnotes. That's a mistake that costs real money. Every time y

A
Agency Script Editorial
June 1, 2026·10 min read
General

Rolling Out AI Hallucinations Across a Team

Most teams discover AI hallucinations the hard way — a confident-sounding wrong answer makes it into a client deliverable, a legal brief, or a published report. The damage isn't just to the output; it

A
Agency Script Editorial
June 1, 2026·11 min read

Ready to certify your AI capability?

Join the professionals building governed, repeatable AI delivery systems.

Explore Certification