suggesting-reviewers

name

suggesting-reviewers

description

Use when an editor needs to identify peer reviewers OR an associate editor (AE) to handle a manuscript from its abstract — matching the paper's topics and methodology to editorial-board members, associate editors, and external experts and excluding ineligible people. Trigger phrases "suggest reviewers", "find reviewers for this paper", "who could review this", "recommend referees", "reviewer candidates", "suggest an AE", "assign an associate editor", "who should handle this paper", "recommend an AE", "select an associate editor". Do NOT use for writing or revising the manuscript — use `academic-research:critic-loop` / `manuscript-revision`. Do NOT use to verify a draft's citations — use `academic-research:fact-check`.

Suggesting reviewers or associate editors

Given a manuscript abstract, propose either peer reviewers or an associate editor (AE) to handle the paper, and explain each candidate's fit. Default journal: Organizational Research Methods (ORM).

Mode — reviewer vs. AE

Detect the mode from the request and announce which one you are running.

• Reviewer mode (default — "suggest reviewers", "find referees", …). Two pools: editorial-board members (filtered by eligibility, matched against pre-built profiles) and external experts (found live from the abstract's topics).
• AE mode ("suggest an AE", "assign an associate editor", "who should handle this paper", …). One pool: the journal's current associate editors. No external search — an AE must be a sitting board officer.

The two modes share everything below except the two steps that explicitly branch (Step 2 eligibility, Step 4 candidate lists). If the request is ambiguous, ask which is wanted before proceeding.

Core principle — combine knowledge with retrieval

Training knowledge is a useful starting point: it maps the field and knows established scholars. But it has two blind spots — lesser-known / junior researchers and very recent work — and those are exactly where this skill adds value. So every run must augment training recall with retrieval: the pre-built board profiles plus live OpenAlex / Semantic Scholar searches that deliberately reach for recent and less-visible names.

Ground each final suggestion in a retrieved publication record wherever one exists — and especially for the less-famous names, where training recall is least reliable. Never put a junior or unfamiliar person in the output on memory alone.

Step 1 — Load the roster and check staleness

The bundled ORM data lives under the plugin root:

${CLAUDE_PLUGIN_ROOT}/skills/suggesting-reviewers/rosters/orm/
  index.md            # membership + role + eligible: yes/no, with snapshot_date
  profiles/<surname>.md

Read index.md. Report to the user the snapshot_date (membership) and the oldest profile checked date. If either is older than 3 months, warn with the age and ask whether to refresh before continuing — do not silently re-fetch. (Refresh procedure: Step 5.)

For a journal other than ORM, ask the user to point at (or build) the equivalent rosters/<journal>/ directory.

Step 2 — Apply eligibility (deterministic)

Reviewer mode.

• Exclude every member tagged eligible: no — current Editor(s)-in-Chief and Associate Editors. They assign reviews; they are never suggested.
• Eligible pool = every member tagged eligible: yes (past editors still listed, Editorial Review Board members, advisory roles).

If an excluded editor is the obvious topical match, do not suggest them — note the exclusion in a transparency line at the end instead.

AE mode. The pool inverts: candidates are exactly the members whose role is Associate Editor. Editors-in-Chief, the Outreach Officer, past/founding editors, and plain Editorial Board members are not AE candidates — only sitting AEs assign and handle papers. (If a specific AE is the submitting author or is otherwise conflicted and the user supplied that information, drop them and note it; otherwise leave conflicts to the assigner.)

Step 3 — Read the abstract → methodology + topics

Classify the paper's methodological orientation — quantitative, qualitative, mixed, or methods-development — and extract its substantive topics. This drives a hard methods filter:

• A clearly quantitative paper surfaces no qualitative-only scholar.
• A clearly qualitative paper surfaces no quantitative-only scholar.
• mixed-or-methodologist profiles match either.

When the paper itself is genuinely mixed, both orientations are in scope.

Step 4 — Build the candidate list(s)

AE mode — read this first. Build a single list from the AE pool of Step 2 and skip the external search entirely (an AE must be a sitting board officer). Match and ground each AE exactly as Step 4 describes for board reviewers below — methods filter (Step 3), topical fit against the profile's topics / representative method papers / ## ORM publications, drilling into mcp__openalex__get_author_works only when a profile is thin or stale. AE profiles are built the same way as ERB profiles (see "Full profile build" under Step 5/Refresh); if a sitting AE has no profile yet, build one on first use and commit it. Then go to Step 5 (AE output). The career-stage tilt and the rest of this step apply to reviewer mode.

Career-stage tilt (both pools): rank credible early-to-mid-career scholars first — these are the people the editor is least likely to already know — but still include a couple of senior options. Label every candidate's career stage. Apply a competence floor: roughly ≥3 relevant papers in the last ~5 years so the person can actually referee the work.

Board reviewers. From the eligible, methods-matched pool, rank by topical fit against each profile's topics, its representative method papers, and its ## ORM publications list (the member's own first-party ORM articles — strong evidence they referee that journal's style well). The profile is the grounding; drill into mcp__openalex__get_author_works or mcp__semantic-scholar__authors-search only if a profile looks thin or stale.

External reviewers (live). From the abstract's topics, search:

• mcp__openalex__search_authors_by_expertise — experts by research area.
• mcp__openalex__search_works (sort=publication_year:desc, from_year recent) — then extract authors, to catch very recent relevant work.
• mcp__semantic-scholar__authors-search — cross-check / fill gaps.

Favor lower-seniority but topically prolific authors (lower works_count / more recent first publication, still active). If the user supplied the paper's authors, do not suggest them; otherwise no COI screening (the editor handles conflicts).

Step 5 — Output

Reviewer mode. Two ranked markdown tables — Editorial board and External — columns:

| Name | Affiliation | ERB role / External | Career stage | Methods | Representative method paper(s) | Fit |

Fit ties the suggestion to specific themes in the abstract. Rank by fit within the junior tilt. After the tables, add any transparency note for a strong-but-ineligible board match, and state explicitly which suggestions are grounded in a retrieved record vs. flagged as training-recall only (there should be none of the latter for junior/unfamiliar names).

AE mode. One ranked markdown table — Associate editors — same columns (the role column reads "Associate Editor"), ranked by topical/methods fit. No junior tilt and no external/transparency lines for ineligible board matches (every candidate is a sitting AE). Still state which suggestions are grounded in a retrieved record.

Refreshing the roster and profiles

Profiles are a hybrid: training knowledge about each scholar, augmented and corrected by retrieved OpenAlex data. Refresh is committed to the repo so both machines (and any ORM AEs) share the same curated data.

Membership diff — WebFetch source_url from index.md, diff against the member list: new eligible members → full profile build (below); departed members → remove their profile; role/eligibility changes → edit index.md. Update snapshot_date.

Primary: one ORM-journal sweep (cheap, every refresh). Collect every ORM article authored by any current member in a single call:

mcp__openalex__collect_works(
    author_ids = [<all eligible-member openalex_id values>],   # any length; chunked by 100 internally
    source_ids = ["S133599136"],                               # ORM
    sort       = "publication_year:desc",
    max_results = 2000)                                          # verify meta.truncated == false

Each returned work carries author OpenAlex IDs, so ID-join to members locally (no name matching). For each member, rewrite the ## ORM publications section (keep article/review, drop editorial/paratext), set orm_works_count, and bump orm_swept. New ORM papers since the last sweep are the cheap signal that a profile's representative papers / orientation may need a look. This one call replaces the old per-member ORM polling.

Occasional: per-member all-venue deep refresh. The ORM sweep only sees ORM papers; a member's defining method papers often sit in other journals. When a profile looks thin or stale (or its ORM count jumped), refresh that member's broader record: mcp__openalex__get_author_works(author_id, from_year=<year of works_through>, sort=publication_year, per_page=200) returns only works since the watermark. If the slice has new method-relevant papers, refresh the representative papers / orientation / specialties / stage and set a new works_through; otherwise just bump checked. Do this for a rotating subset, not all 129 every time.

Full profile build (per member).

1. Resolve to an OpenAlex author: mcp__openalex__search_authors(query=name). Do not combine the institution and exact_phrase parameters — that returns HTTP 400; pass the name alone and disambiguate from the results by last_known_institutions + topical plausibility + ORCID. Pick the entry whose affiliation matches the board listing and that has a real publication record (not a 1–3 work near-empty namesake). Store openalex_id/orcid. If ambiguous, flag for manual confirmation — never guess an ID.
2. mcp__openalex__get_author_profile(author_id, top_works_count=10, recent_works_count=8) → h-index, citations, main topics, top + recent works.
3. Select representative method papers — those showing a methodological contribution (technique, measurement, analytic critique), regardless of journal (a method paper in a substantive journal counts). Ignore non-article clutter in recent_works: figshare/OSF "Additional file", datasets, IRB materials, and other type: other/repository entries are not papers — filter to real article/review/book-chapter items with a journal source.
4. Classify methods_orientation; derive career_stage from earliest affiliation/work year + works_count + h-index, augmented with training knowledge.
5. Write the profile with built, works_through (newest real publication), checked dates.

Profile file shape

name: <Full Name>
openalex_id: A123456789
orcid: 0000-0000-0000-0000      # if available
role: <ERB role from index.md>
eligible: yes
methods_orientation: quantitative | qualitative | mixed-or-methodologist
specialties: [SEM, measurement invariance, ...]
career_stage: early | mid | senior
career_signals: { first_pub_year: YYYY, works_count: N, h_index: H }
built: YYYY-MM-DD
works_through: YYYY-MM-DD        # newest all-venue publication considered (deep-refresh watermark)
checked: YYYY-MM-DD              # last refresh check, even if nothing changed
orm_works_count: N              # first-party ORM articles found in the last journal sweep
orm_swept: YYYY-MM-DD           # date of that ORM-journal collect_works sweep

Body: representative method papers (one line each on the contribution), the topics the member would referee well, short training-knowledge notes the data doesn't capture, and a ## ORM publications section (the member's own ORM articles, newest first) written by the journal sweep.

Red flags — stop

• Reviewer mode: about to suggest a current EIC/AE (eligible: no) → drop; they're excluded.
• AE mode: about to suggest a non-AE (EIC, Outreach Officer, past/founding editor, plain Editorial Board member) → drop; only sitting AEs qualify.
• AE mode: running an external search or applying the junior tilt → stop; AE mode is the AE pool only, ranked by fit.
• A junior/unfamiliar name with no retrieved record behind it → drop or ground.
• An opposite-orientation scholar slipping past the methods filter → remove.
• A candidate below the competence floor → remove.
• Relying on training recall without running the augmenting searches → go back to Step 4 and search.