/mstack:lit-map

Stage: map · Voice: systematic-reviewer

After /mstack:research-question returns green or yellow. One question: what conversation is this paper joining, and where does it crack the consensus?

Procedure

1. Load .mstack/research-question.md; the contribution sentence is the seed. If lit/index.md exists, read it first: triaging the converted corpus costs a fraction of opening papers. If lit/pdf/ holds unconverted PDFs, run /mstack:pdf-ingest before going further; searching the web for abstracts of papers you already own is backwards.
2. Define the search in one paragraph and get sign-off: 3–5 keyword combinations; the 2–3 most directly engaged authors; time window (typically 10 years plus the foundational older work); venues (top-3 field journals plus SSRN, NBER, OSF).
3. Search. Local corpus first (lit/index.md, then the relevant section of a converted paper rather than the whole file). WebSearch for keyword + author scans; WebFetch (or a Google Scholar / Semantic Scholar MCP if available) for abstracts and citation counts; snowball backward (the reference lists of the Foundation papers) and forward ("cited by" on the two or three papers closest to the claim). Check for replications, meta-analyses, and many-analysts studies of the headline effect, and for working papers on SSRN, NBER, OSF, and conference programs (APSA, MPSA, EPSA, IPES) that a published-only search misses. Record the search strings and dates in the map so the search is reproducible. Build 20–40 candidates.
4. Triage each paper: Foundation (3–5; the canonical works the conversation rests on; must-cite), Frontier (5–10; recent work in active conversation with the claim; engage explicitly), Adjacent (5–10; related but a different question; cite once), Discard (noted in the file, not engaged).
5. Map the consensus (Foundation + Frontier, one paragraph answering three questions):
   • What does the consensus believe?
   • Where does it crack; what is contested?
   • What does the canonical work not answer?
6. Locate the gap (one paragraph answering three questions):
   • Which paper(s) does the project most directly engage?
   • What does it add that they lack?
   • Which hostile reviewer would object, and why? Name the two or three scholars most likely to referee the paper and what each would insist on; /mstack:referee-mock uses this.
7. Stage refs.bib. A BibTeX entry per Foundation + Frontier paper, appended to paper/refs.bib (dedupe by key, never overwrite). Do not invent fields; leave unknowns blank with a % TODO. Prefer entries from lit/refs-ingested.bib for ingested papers, but only those without a % TODO verify marker (metadata parsed from the PDF rather than fetched from Crossref).
8. Save to .mstack/lit-map.md: classification, synthesis, gap. Set paper.status: "mapping" in .mstack/config.yaml if it still says ideating.

Outputs

• .mstack/lit-map.md — map, consensus, gap statement.
• paper/refs.bib — appended entries.
• Summary block: top 5 must-engage papers, the gap, next skill.

Anti-patterns

• A list instead of a synthesis. Thirty papers is not a map; the synthesis is.
• Inventing citations. Not found by search means left out.
• Too broad a frontier. Five engaged papers beat fifteen name-drops.
• Hiding contradictions. If the consensus contradicts the claim, say so now rather than at R1.

Next

/mstack:theory-build, then /mstack:hypothesis-design and /mstack:identification-review. A lit map is a prerequisite for drafting intro or theory.