econ-research

Plugin econ-research from marketplace econ-research-ai-kit in k-mizrahi/econ-research-ai-kit. Commands read from the repo's manifest — 3 days ago.

Skills (23)

• add-exemplars

  Setup step for the writing and slide skills — run once per user, before the first draft. Builds private notes on papers the user likes for their writing, talks they like, and the user's own past slide decks, then a style profile that every write-* skill reads before drafting and that write-slides us

• blast-radius

  Consumer sweep after a data-layer bug is found — enumerate every reader of the affected file/column, patch the generating script, and list stale outputs. Use PROACTIVELY whenever a bug is found in a shared data file, panel column, or classification — before fixing only the call site where it was dis

• check-output

  Verify that numbers cited in a draft (paper, slides, notes) match current output files — catches stale results before submission

• design-check

  Mechanical-bias preflight for any NEW sample screen, outcome variable, treatment definition, or timing rule — does the definition build an effect in by construction? Run BEFORE adopting it and re-running the chain. Use PROACTIVELY when the user proposes a new screen/outcome/definition; this is the m

• explore-directions

  Explore new directions for a paper end-to-end. Agents generate and screen ideas; the user picks; each pick is pre-registered (kill condition red-teamed), implemented in its own git worktree, convention-audited, and judged by a seminar panel; a PDF report summarizes every idea, its results, and the v

• handoff

  Write a durable end-of-session handoff document recording where things stand, what changed, findings worth remembering, and ordered next steps. Use ONLY when the user explicitly asks for a handoff (e.g. "handoff", "write a handoff", "hand off this session") — never trigger automatically.

• log-decision

  Record a ratified or provisional methodological decision in the project's routed decision register (as the project's CLAUDE.md names it) and run the enforcement sweep in the same session. Use PROACTIVELY whenever the user ratifies a decision ("let's adopt X", "drop Y", "make Z the headline") or a sc

• log-result

  Append a result entry to the project's routed results log (as the project's CLAUDE.md names it) — called automatically after any analysis produces new findings

• project-summary

  Compile an audience-facing project orientation document — what the project is, why it matters, current results of record, binding decisions, gated claims, and open work — from the project's state-of-record docs, so the user or a co-author can catch up before moving to next steps. Use when asked for

• read-paper

  Read and summarize an academic paper (PDF). Use PROACTIVELY when the user references a paper or methodology you haven't read yet — do not wait to be asked.

• refresh

  Run a full-project audit and reconcile the defining documents (CLAUDE.md, PURPOSE.md, key-decisions doc, plans) and memory with reality — detect drift, surface contradictions, record unrecorded decisions, promote durable memory into docs, prune stale memory. Interactive and heavy by design. Use ONLY

• seminar

  Experimental multi-agent research feedback — simulated reactions for stress-testing an argument, not a substitute for real referees or the researcher's judgment. PANEL mode (default) simulates a faculty seminar — two field experts (personas built for the paper's fields) and an outside-field economis

• setup-project

  Set up the research-workflow scaffolding in the current project — a CLAUDE.md in the standard section order, the plans rule, a key-decisions register, results log, dropped-analyses doc, backlog, handoffs/, a skill-bindings file, the user's writing-style profile if none exists yet, and optionally the

• sync-state

  End-of-session pointer reconciliation — sync the repository's state-of-record documents so the next session starts from correct pointers; optionally mirror headline changes into persistent agent memory

• triage-feedback

  Categorize external feedback (referee reports, advisor comments, seminar Q&A) into actionable / parkable / inherent / impossible / addressed

• write-data

  Draft or revise a paper's data section — sources, sample construction, linkages, variable definitions, summary stats, limitations — using the canon's construction-narration playbook and the papers the user likes. Use when the user asks to write or tighten a data/setting section.

• write-design

  Draft or revise a paper's research-design / empirical-strategy section — estimating equations, identifying assumptions, threats, event-study setup — using the canon's identification-exposition playbook and the papers the user likes. Use when the user asks to write or tighten an empirical strategy, i

• write-intro

  Draft or revise a paper introduction (or abstract) using the econ-writing playbook — Head's formula, the contract test, and the intro architectures of papers the user likes. Use when the user asks to write, outline, or tighten an introduction or abstract.

• write-litreview

  Write or revise the literature-positioning material of a paper — the closest-papers and contribution paragraphs at the end of the intro, with an optional Related Literature title — using the canon's non-combative contribution grammar. Use when the user asks for a lit review, related-literature secti

• write-onepager

  Compile a one-page extended abstract plus 3–4 figure pages for one paper — question, motivation, design, main results, mechanisms in brief, status and open decisions — written in the readable register of a top-5 introduction, for an outside reader. Compiles from a fixed manifest of state-of-record s

• write-plan

  Draft a dated, living plan for the next piece of work — from a backlog item, a new idea, or feedback — checked against the decisions register and dropped analyses, with phases, expected outputs, decision rules written before any results, and a Results section that grows as the work proceeds. Use whe

• write-results

  Draft or revise a paper's results section — table/figure walk-throughs, magnitude conversion, null phrasing, robustness compression, mechanisms — using the canon's results-narration playbook and the papers the user likes. Use when the user asks to write or tighten results, findings, or mechanisms te

• write-slides

  Build or revise a beamer slide deck for a research talk — opening contract, claim titles, figure/table-first minimal-text frames, concern→test robustness, wired backup armor — plus the mandatory companion per-slide notes PDF, using the slide-craft playbook from Shapiro and Goldsmith-Pinkham and the

Similar plugins