Sources to a cited brief

# Question
Should the town library move to a four-day opening week?

# Source A: Library usage report, 2025 (excerpt)
Visits fell 12% between 2023 and 2025. Tuesday and Wednesday are the busiest days, with 38% of weekly visits combined. Monday is the quietest day, with 6% of weekly visits. The report does not cover online borrowing.

# Source B: Staff survey summary, March 2026 (excerpt)
Of 14 staff, 9 said they would prefer a four-day week. Several mentioned fewer sick days as a likely benefit. Two staff said they were worried about longer shifts. The survey was voluntary and 14 of 17 staff replied.

# Source C: Council budget note, April 2026 (excerpt)
Closing one day a week would save an estimated 18,000 EUR per year in heating and casual staff costs. The note says the estimate has not been checked by finance. No figure is given for lost fines or room hire income.

# Source D: Reader comment letter, May 2026 (excerpt)
A group of parents asks that Saturday stays open because school-age children use the library at weekends. The letter has 41 signatures. It does not say which weekday could close instead.

# Research brief: four-day opening week for the town library

Question: Should the town library move to a four-day opening week?
Sources used: A (usage report 2025), B (staff survey, March 2026), C (council budget note, April 2026), D (reader letter, May 2026). Nothing else.

## Bottom line
The pasted sources do not settle the question. They suggest Monday is the cheapest day to close and that staff lean in favour, but the saving is unchecked and weekend use by families is a concern. Confidence: low to medium.

## Findings
1. Monday is the quietest day, with 6% of weekly visits; Tuesday and Wednesday together have 38%. [A] Confidence: high (stated figures in one report).
2. Overall visits fell 12% between 2023 and 2025. [A] Confidence: high. The report does not cover online borrowing, so this is only about visits.
3. 9 of 14 staff who answered prefer a four-day week; 14 of 17 staff replied; two worried about longer shifts. [B] Confidence: medium (voluntary survey, small group).
4. Staff expect fewer sick days, but no figures are given. [B] Confidence: low (opinion, no data).
5. Closing one day a week is estimated to save 18,000 EUR a year. [C] Confidence: low (the note says finance has not checked it, and lost fines and room hire income are not counted).
6. 41 parents signed a request to keep Saturday open for school-age children. [D] Confidence: medium (clear request, but a single group's view).

## Where sources agree or pull apart
- A and C point the same way on cost and demand: a low-use day exists and closing a day saves money. Together they support Monday as a candidate, but no source says "close Monday".
- B favours fewer days; D asks to protect one specific day. They are not in conflict unless the closed day is Saturday, and D does not name an alternative.

## Open questions
- Has finance checked the 18,000 EUR estimate, and what are the lost fines and room hire income? (C is silent.)
- How would online borrowing change the picture? (A excludes it.)
- Which weekday would the 9 staff prefer to close, and how long would shifts become? (B gives no detail.)
- Which day would the parents accept closed? (D does not say.)

## Not answerable from these sources
- Effect on opening-hours coverage for people who work weekdays.
- Whether other towns saw visits change after cutting a day.

---
name: research-brief
description: Turns a few pasted source excerpts and a question into a short brief with citations to those sources only, a confidence level per finding and a list of open questions. Use when someone needs an answer grounded strictly in the material they pasted.
---
# Research brief

A question plus pasted sources in, a cited brief with confidence levels out.

## When to use
- The user pastes 2 to 6 excerpts (reports, notes, letters, articles) and a question.
- They need to know what the material supports, and what it does not.

## Inputs
- The question, in one sentence. If missing, ask for it.
- The excerpts as text or a file path. Label each source A, B, C... using the title the user gave; if there is no title, ask.
- Optional: the audience and a length limit.

## Steps
1. Read every excerpt. Note for each: what it claims, any figures, its date, its limits (what it says it does not cover).
2. Split the material into single findings. One finding is one claim with one citation such as [A].
3. Give each finding a confidence level:
   - high: stated plainly with figures or clear wording in the source, and not contradicted
   - medium: a clear statement but from a small, voluntary or one-sided group
   - low: opinion, estimate the source itself calls unchecked, or no data
4. Compare sources: where they agree, where they differ, and where they only seem to conflict.
5. Write a bottom line of two or three sentences that answers the question only as far as the findings allow. Say plainly when they do not settle it.
6. List open questions: what the sources leave out, what a source says it did not cover, and what one more source could answer.
7. List what is not answerable from these sources.

## Output
Write `research-brief.md` as a new file with: title, the question, the sources used, `## Bottom line`, `## Findings` (numbered, each with citation and confidence), `## Where sources agree or pull apart`, `## Open questions`, `## Not answerable from these sources`.

## Rules
- Never invent facts, figures or quotes. Cite only the pasted sources, with [A], [B] style tags.
- Do not use outside knowledge as if it were in the sources. If you add context, put it under "Not answerable" as a question, never as a finding.
- Write results to a new file, never overwrite the pasted material.
- Keep numbers exactly as the sources state them; do not round or combine unless you show the sum.
- Do not make recommendations beyond what the findings support.
- Keep the material local; do not search the web.
- Out of scope: legal, medical, tax or financial advice. Flag such items for a professional.