theme-lit

# Theme Lit

An annotated bibliography is not a review. Write 3-6 theme sentences that argue what the field knows, where it splits, how strong the proof is, and what is still open.

Inspired by SkillMedev academic-researcher literature-review (MIT). Original Grok remake. Complements claim-line (argumentative essay spine with thesis/CEA, not a review), study-loop (flashcards), quiz-bank, score-grid. Education only. No fake citations.

Read `references/review-rails.md` before you write. Fill the three templates under `templates/`.

## When to use

- Related-work section inside a paper, proposal, or thesis chapter
- "What does the research say?" that should become a synthesis, not a source dump
- A pile of papers that need grouping by argument, not by author or year
- A named gap that points at the student's own question
- Drafting a review without inventing authors, titles, years, DOIs, pages, or quotes

## When NOT to use

- Systematic / PRISMA protocol review (search strings, inclusion counts, risk-of-bias tables, completeness claims) → out of scope. Tell them they need a protocol. Do not fake exhaustiveness.
- Thesis-driven argumentative essay that is not a review → use claim-line
- Exam flashcards → use study-loop
- Quiz / item bank → use quiz-bank
- Analytic rubric for the paper → use score-grid
- Inventing a bibliography, placeholder authors, fake DOIs, or made-up quotes

## Inputs

- Question or topic the review must serve (required). If they have a project question, lock it; the gap must point at it.
- Audience and job: related-work section vs standalone review vs undergraduate paper
- Length target if they named one. Related-work sketch ~800-1500 words; standalone longer. Label guesses.
- Sources they already have (titles, PDFs, notes). Treat only what they pasted as known. Everything else is a `[CITE: type]` slot.
- Grade band or course. Age-appropriate language.
- Date window, languages, or study types if they named an inclusion posture. Representative, not exhaustive.

## Procedure

1. Read `references/review-rails.md`. If they asked for a systematic/PRISMA protocol, stop. Say they need a protocol this skill does not run. If they asked for a thesis-driven essay that is not a review, hand off to claim-line. If they want flashcards, hand off to study-loop.
2. Fill `templates/scope.md`. Lock the question, audience, length (label guesses), corpus they actually have, and inclusion posture. Never claim the set is complete.
3. For each work they supplied, extract claim / method / setting / limitation. Do not stop at "Author found X."
4. Group by theme sentences, not authors, not chronology, not topic nouns. Draft 3-6 sentences that argue a slice of the field. Fewer than 3 is a summary of two papers. More than 6 is a topic list wearing a costume.
5. Fill `templates/themes.md`. Every theme states consensus, disputes (name both sides and a plausible why), and evidence quality (rigor, not volume; echoes of one study are not extra evidence). Each theme puts at least two works in conversation.
6. Name the gap. Unanswered question, unstudied population, unapplied method, or a thin-evidence claim the field keeps repeating. If this review supports their project, the gap must point at their question.
7. Fill `templates/gap-cites.md`. Mark every missing source as `[CITE: type]`. Flag unverifiable items they mentioned but did not produce. Never fabricate authors, titles, years, DOIs, pages, or quotes.
8. Sketch length: related-work ~800-1500 words unless they gave a number; standalone longer. Label the target a guess when they did not specify.
9. Growing the set: start from a field-defining work plus a few recent ones they already have; walk a handful of central references backward (who those papers cite) and forward (who cites them). Representative, not a database dump. Hand citation formatting and DOI checks to a future cite skill.
10. Deliver the three templates. One closing line: this is a thematic argument about the state of knowledge, not a systematic review and not an essay spine.

## Output format

- Scope card (question, audience, length, corpus, inclusion posture, labeled guesses)
- Theme sheet (3-6 theme sentences; consensus / disputes / quality per theme; at-least-two-works rule)
- Gap and cite card (named gap, how it feeds their question, `[CITE]` slots, unverifiable flags)
- One line: thematic review argument; not a PRISMA protocol and not a claim-line essay

## Safety

- Follow `references/review-rails.md` with no exceptions.
- NEVER fabricate authors, titles, years, DOIs, pages, quotes, or data. Use `[CITE: type]` markers. Flag unverifiable items.
- No student PII. No live exam keys. No CSAM. No sexual content involving minors. No self-harm examples.
- Age-appropriate for the stated grade band.

## Stop conditions

- Three templates delivered, or refused for systematic/PRISMA protocol / claim-line essay / study-loop flashcards / invented bibliography / missing question