PP11: The Research and Analysis Pack

You are helping me turn a broad topic into one research question I can answer, and telling me what evidence to go and get.

TOPIC OR STARTING QUESTION:
[PASTE]

THE DECISION THIS RESEARCH HAS TO SUPPORT:
[PASTE, one line]

WHO READS THE FINAL OUTPUT:
[PASTE, one line]

TIME AND EFFORT AVAILABLE:
[PASTE, one line]

SOURCES OR LIMITS I HAVE TO WORK WITHIN:
[PASTE, or write 'none']

MARKETS, SECTORS OR JURISDICTIONS THAT MATTER:
[PASTE, one line]

WHAT TO DO:
1. State the core research question in one sentence.
2. Give the three sub-questions that have to be answered before the core question can be.
3. Define the scope: what is in, what is out, over what period, in which markets.
4. For each sub-question, list the source types that would answer it, strongest first. Say what each type can settle and what it cannot.
5. Build the source plan. For each source type give me the exact search terms to use, the places to run them (name the databases, official statistics bodies, regulators, industry bodies, company filings, trade press or archives that hold this kind of material), and what to record when I find something.
6. State the test for a useful answer: what I would need to see before I could act on it.
7. List the assumptions built into my framing that need checking.

THEN GIVE ME:
- The one sub-question that most changes the answer if I get it wrong.
- The source type I am most likely to skip and most likely to regret skipping.
- Two questions I need to answer to sharpen the frame.

CONSTRAINTS:
- Plain English. No hype. No flattery. No em-dashes.
- Do not invent sources, figures or quotations.
- Name categories and search terms freely, but mark any specific publication, dataset or institution as unverified until I have opened it myself.

OUTPUT:
- The research frame: core question, three sub-questions, scope in and out, assumptions to check.
- The source plan: ranked source types per sub-question, search terms, named places to look, what to record.
- The test for a useful answer, and the two sharpening questions.

Why it works: A research question you cannot answer with the sources available to you is a wish, not a plan. Making the model name the search terms and the places to look turns the frame into a list you can work through.

You are helping me triangulate the sources below on the same question. Work only from the extracts I have pasted.

RESEARCH QUESTION:
[PASTE, from the question frame]

SOURCE A:
[PASTE THE RELEVANT EXTRACT OR FIGURES]
Publisher, author, date and URL: [PASTE]

SOURCE B:
[PASTE THE RELEVANT EXTRACT OR FIGURES]
Publisher, author, date and URL: [PASTE]

SOURCE C:
[PASTE THE RELEVANT EXTRACT OR FIGURES]
Publisher, author, date and URL: [PASTE]

FURTHER SOURCES IN THE SAME FORMAT:
[PASTE, or write 'none']

WHAT I AM USING THESE SOURCES TO SUPPORT:
[PASTE, one line]

WHAT TO DO:
1. Summarise the core finding of each source in one sentence.
2. Set out where all the sources agree.
3. Set out where they disagree, and state what evidence would settle each disagreement.
4. Rate each source on evidence type, independence, recency and how clearly it states its method. Give the reason for each rating.
5. Name any point carried by only one source, since that is the thinnest part of the picture.
6. Flag any two sources that share an origin, a funder or a dataset, because agreement between them is not independent confirmation.
7. Name what is missing, and which source type would fill the hole.

THEN GIVE ME:
- The finding I can state with most confidence, and the wording I should use.
- The finding I should not yet state at all.
- Two questions I need to answer to sharpen the comparison.

CONSTRAINTS:
- Plain English. No hype. No flattery. No em-dashes.
- Work only from the extracts I pasted. Do not add sources, figures or quotations of your own.
- Do not assume the contents of a link I have not quoted.

OUTPUT:
- A source-by-source comparison with agreement, disagreement and credibility ratings with reasons.
- Single-source and shared-origin flags, and the gap in the evidence.
- The safe-to-state finding, the not-yet finding, and the two sharpening questions.

Why it works: Two sources agreeing means nothing if they draw on the same dataset or the same funder. Asking for shared origins and for points carried by a single source separates real corroboration from an echo.

You are helping me triage a draft for the three claims most worth verifying.

DRAFT:
[PASTE DRAFT]

CLAIMS ALREADY CHECKED:
[PASTE, or write 'none']

FOR EACH OF THE THREE CLAIMS YOU FLAG:
1. State the claim verbatim.
2. Explain why (high stakes, surprising, unsourced).
3. Recommend the source type that would settle it.
4. Suggest one specific source.
5. Predict whether the claim will hold up, and state what your prediction is based on.

THEN GIVE ME:
- The one claim that, if false, would cost the most credibility.
- The one claim where the cited source likely does not say what the draft says.
- Two questions I need to answer to sharpen the triage.

CONSTRAINTS:
- Plain English. No hype. No flattery. No em-dashes.
- Any source you name is a lead for me to check, not a citation, so mark it unverified.

OUTPUT:
- The three claims with rationale, source type, source suggestion and a prediction with its basis.
- The highest-risk claim, the most likely misreading of a source, and the two sharpening questions.

Why it works: Verification budgets are small. Ranking by consequence rather than by doubt puts the effort where a mistake would actually damage you, which is rarely where the doubt feels strongest.

You are helping me judge whether the method below supports the claim being made of it.

METHOD SECTION OR DESCRIPTION OF THE STUDY:
[PASTE]

THE CLAIM THIS STUDY IS BEING USED TO SUPPORT:
[PASTE]

MY FIELD AND THE COMPARISON THAT MATTERS TO ME:
[PASTE, one line]

WHAT TO DO:
1. Restate the method in plain English in three sentences or fewer.
2. State the sample: size, how it was selected, who is left out, over what period.
3. List the design limits: bias risk, missing control or comparator, confounders, and anything measured by proxy rather than directly.
4. State what this method can show, and what it cannot.
5. Judge whether the stated conclusion goes further than the method allows. Quote the sentence where it does.
6. Say how this finding would most likely fail to replicate.

THEN GIVE ME:
- The weakest single point in the method.
- The one extra piece of information that would most change my read.
- Two questions I need to answer to sharpen the check.

CONSTRAINTS:
- Plain English. No hype. No flattery. No em-dashes.
- Work only from the method I pasted. Do not assume detail that is not there, say it is missing.

OUTPUT:
- The plain-English restatement, the sample, and the design limits.
- What the method can and cannot show, any overreach quoted, and the replication risk.
- The weakest point, the information most worth having, and the two sharpening questions.

Why it works: Asking the model to quote the sentence where the conclusion overreaches makes the criticism checkable. If it cannot quote the sentence, it has not found the problem.

You are helping me find the strongest case against my claim.

MY CLAIM OR DRAFT PASSAGE:
[PASTE]

WHO WILL READ OR HEAR THIS:
[PASTE, one line]

OBJECTIONS I HAVE ALREADY HEARD:
[PASTE, or write 'none']

WHAT TO DO:
1. State the strongest counter-argument in its best form, as the most capable person holding it would put it.
2. Give the evidence behind it. Label anything you cannot verify from what I pasted.
3. Name the weakest link in my claim, quoting my own words.
4. Say whether the counter-argument changes my conclusion or only how I should word it.
5. State what evidence would settle the disagreement.
6. Rewrite my claim in one sentence so it survives the counter-argument without hedging.

THEN GIVE ME:
- The objection most likely to be raised in public.
- The objection most likely to be right.
- Two questions I need to answer to sharpen my position.

CONSTRAINTS:
- Plain English. No hype. No flattery. No em-dashes.
- Do not soften the counter-argument to protect my position.
- Do not invent evidence for either side.

OUTPUT:
- The strongest opposing case with its evidence and verification labels.
- The weakest link quoted from my draft, and the effect on my conclusion.
- The rewritten claim, and the two sharpening questions.

Why it works: Separating an objection that changes your conclusion from one that changes only your wording stops you rewriting a sound argument for presentational reasons, and stops you defending an unsound one.

You are answering as an experienced practitioner in the role and discipline named below. Do not impersonate a named individual and do not invent credentials or career history.

TOPIC:
[PASTE]

THE PRACTITIONER ROLE AND DISCIPLINE I WANT (a role, not a named person):
[PASTE]

MY DRAFT OR FINDINGS SO FAR:
[PASTE]

WHAT I WANT OUT OF THE INTERVIEW:
[PASTE, one line]

WHAT TO DO:
1. Give the three things people new to this topic most often get wrong.
2. Name the angle a practitioner would check first that my draft has missed.
3. Ask me the five questions a practitioner would want answered before accepting my findings.
4. Point out any term I am using loosely that has a precise meaning in this field, and give the precise meaning.
5. Name the source types a practitioner would trust here, and why. If you name a specific work, mark it unverified so I can check it before I use it.

THEN GIVE ME:
- The single correction that would most improve the draft.
- The part of my draft a practitioner would find naive.
- Two questions I need to answer before the next draft.

CONSTRAINTS:
- Plain English. No hype. No flattery. No em-dashes.
- Speak as a role, not as a real named person.
- Any specific work you name is unverified until I check it.

OUTPUT:
- The common errors, the missed angle, and the five practitioner questions.
- Loose terms corrected, and the source types a practitioner trusts.
- The highest-value correction, the naive passage, and the two sharpening questions.

Why it works: A practitioner's first five questions expose more than any summary does. The ones you cannot answer are the gaps in your work, and you find them before your reader does.

You are helping me map the literature around the gaps I have already found.

TOPIC:
[PASTE]

TIME FRAME AND ANY DISCIPLINES OR JURISDICTIONS THAT MATTER:
[PASTE, one line]

GAPS AND DISAGREEMENTS FOUND SO FAR:
[PASTE, from prompts 2 to 6]

READING LIST OR SOURCES I ALREADY HOLD:
[PASTE, or write 'none']

WHAT TO DO:
1. Group the field into themes or schools of thought. Give the position each one holds in a single sentence.
2. For each theme, say where the evidence is strong and where it is thin.
3. Name the live disagreement in the field and what it turns on.
4. List candidate works, authors and institutions worth checking. Mark every one unverified unless it appears in the reading list I pasted. Do not give page numbers, quotations or identifiers you cannot verify.
5. Give the search terms and the databases or source types I should use to confirm each candidate.
6. Name what looks missing from the field, and why that gap might exist.

THEN GIVE ME:
- The one theme most likely to overturn my current answer.
- The three candidates to check first, in order.
- Two questions I need to answer to sharpen the map.

CONSTRAINTS:
- Plain English. No hype. No flattery. No em-dashes.
- Never present an unchecked reference as a citation.
- If you are not confident a work exists as described, say so and give the search terms instead.

OUTPUT:
- Themes with positions, evidence strength per theme, and the live disagreement.
- A candidate list with unverified labels, plus search terms and databases for each.
- The theme that most threatens my answer, the ranked three to check first, and the two sharpening questions.

Why it works: A model with no web access will produce references that look right and are not. Forcing an unverified label on every name, and search terms instead of identifiers, turns a fabrication risk into a to-do list.

You are drafting a synthesis from research I have already done. Use only the material I pasted. Do not add facts, figures or sources of your own.

RESEARCH FRAME AND SOURCE PLAN:
[PASTE, prompt 1]

TRIANGULATION:
[PASTE, prompt 2]

CLAIM TRIAGE AND VERIFICATION RESULTS:
[PASTE, prompt 3]

METHOD CHECKS:
[PASTE, prompt 4]

COUNTER-ARGUMENTS:
[PASTE, prompt 5]

PRACTITIONER REVIEW:
[PASTE, prompt 6]

LITERATURE MAP:
[PASTE, prompt 7]

READER, FORMAT AND LENGTH:
[PASTE, one line]

WHAT TO DO:
1. Answer the core research question in one sentence.
2. Write the executive summary in under 150 words.
3. Structure the body around the three sub-questions from the research frame.
4. Sort every point into confirmed, likely, or open. Say which source or check puts it there.
5. State the strongest counter-argument and how the answer survives it, or where it does not.
6. List what still needs verification before this is sent or published, and name the prompt in this pack that will do it.
7. Close with what the reader should do next.

THEN GIVE ME:
- The sentence a hostile reader would attack first.
- The one claim carrying the most weight on the thinnest evidence.
- Two questions I need to answer before this goes out.

CONSTRAINTS:
- Plain English. No hype. No flattery. No em-dashes.
- Use only what I pasted. Where an earlier output ran long and I pasted my own summary of it, work from the summary.
- If a section cannot be written from the material supplied, say what is missing rather than filling the gap.

OUTPUT:
- The one-sentence answer and the executive summary.
- The body on the sub-questions, with every point sorted into confirmed, likely or open and its basis named.
- The counter-argument and the response, the outstanding verification list, next actions, and the two sharpening questions.

Why it works: Most research writing fails by presenting likely as confirmed. Sorting every point and naming what put it there makes the confidence visible, to your reader and to you.