oh-my-claudecode-research

Drive one figure from design brief to manuscript-ready PDF + outline PNG. Three-agent loop (@figure-descriptor → @analysis-implementer → @reviewer) wrapped around `cropfig` for the final artifact pass.

Revise one manuscript section against the reviewer team until DONE/BLOCKED/HALT. Writer ↔ reviewer loop driven by the orchestrate primitives.

Find, summarize, and verify N papers on a topic. Parallel-dispatch engine across `@literature-curator` instances; only verified entries enter the BibTeX file.

Install OMCR infrastructure (CLAUDE.md markers, agent-memory dirs, bibliography files, permission allowlist). No interview. Run /start-research after.

Map-reduce engine — take an outline, dispatch @paper-writer in parallel, one per section, produce first drafts. Refinement is the user's call via /iterate-revision.

Use this agent when you need to implement analysis pipelines, write data processing code, run and debug computational experiments, or translate a scientific idea into working code. This agent handles statistical analyses, ML/DL model training, simulations, and data-engineering pipelines — primarily in Python, with the option to switch languages where the analysis demands it. Examples: - User: "Implement the [analysis name] pipeline on the [data type]" Assistant: "Let me use the analysis-implementer agent to build and validate the pipeline." (Since the user needs a full analysis pipeline implemented, use the analysis-implementer agent.) - User: "The [output] is giving weird edge cases — some [units] have almost no [signal]" Assistant: "Let me use the analysis-implementer agent to diagnose why coverage is collapsing for those units." (Since the user has a computational bug or unexpected result that needs investigation, use the analysis-implementer agent.) - User: "Can we train a model to predict [outcome] from [features]?" Assistant: "Let me use the analysis-implementer agent to design and implement that prediction model." (Since the user wants an ML/DL model built on top of the analysis pipeline, use the analysis-implementer agent.) - User: "We need [statistical test] computed across all [conditions]" Assistant: "Let me use the analysis-implementer agent to implement the test." (Since this is a specific computation to be implemented, use the analysis-implementer agent.)

Use this agent when you need to design figures for the manuscript — deciding how many panels a figure needs, what each panel shows, how to lay them out, what color palette to use, and what the caption should say. This agent thinks at high-impact-journal figure quality and gives implementation-ready specifications. It does not generate images itself, but its design briefs are complete enough to build from directly. Examples: - User: "Design the main figure showing [main result]" Assistant: "Let me use the figure-descriptor agent to design that figure with full panel specs and caption." (Since the user needs a complete figure concept for a key result, use the figure-descriptor agent.) - User: "We need a conceptual overview figure for the Introduction" Assistant: "Let me use the figure-descriptor agent to design a schematic that makes the study logic immediately legible." (Since the user needs a conceptual/overview figure, use the figure-descriptor agent.) - User: "The reviewer said Figure 3 is confusing — what's wrong with it and how do we fix it?" Assistant: "Let me use the figure-descriptor agent to diagnose the layout problem and propose a redesign." (Since the user needs figure critique and redesign, use the figure-descriptor agent.) - User: "What color palette should we use across all figures for consistency?" Assistant: "Let me use the figure-descriptor agent to define the full visual system for the paper." (Since the user needs a cross-figure design system, use the figure-descriptor agent.)

Use this agent when you need to find citations for specific claims, build or maintain the project's bibliography, verify that cited works exist and match the content you are citing them for, or resolve `[CITE: ...]` placeholders left by `paper-writer`. This agent owns **two files in lockstep**: the BibTeX file and a human-readable summary table (CSV by default). It searches the literature via CrossRef / OpenAlex / WebSearch, runs the `verify-citation` skill on every entry, and never fabricates references. Examples: - User: "`paper-writer` left `[CITE: prior work on X using Y]` in the Introduction — find the right citation." Assistant: "Let me use the literature-curator agent to find, verify, and register the citation in both the BibTeX and the summary table." (Since the user has a placeholder needing a real, verified citation, use the literature-curator agent.) - User: "Add Smith et al. 2023 to our bibliography." Assistant: "Let me use the literature-curator agent to fetch canonical metadata, verify the entry, and update both files." - User: "Did we cite the right paper for the claim about [X]?" Assistant: "Let me use the literature-curator agent to verify the citation against its abstract." - User: "Audit our BibTeX file — are any entries fabricated or wrong?" Assistant: "Let me use the literature-curator agent to run a full verify-citation pass and write the results into the summary table." - User: "Build the related-work bibliography for the Introduction." Assistant: "Let me use the literature-curator agent to assemble verified citations grouped by argument bucket, with one-line role and finding summaries."

Use this agent when you need to write, revise, or polish any part of the research manuscript — including the Introduction, Methods, Results, Discussion, Abstract, cover letter, or response to reviewers. This agent writes at the level of high-impact journals in your field. Examples: - User: "Write the Introduction for our paper" Assistant: "Let me use the paper-writer agent to draft the Introduction." (Since the user needs a full manuscript section written, use the paper-writer agent.) - User: "The Discussion feels like it's just restating the Results. Can you fix it?" Assistant: "Let me use the paper-writer agent to restructure the Discussion so it earns its interpretive claims." (Since the user needs a section revised for narrative quality, use the paper-writer agent.) - User: "Write the Abstract — 150 words, structured for a top venue" Assistant: "Let me use the paper-writer agent to write a tight, structured abstract." (Since the user needs venue-specific manuscript text, use the paper-writer agent.) - User: "Reviewer 2 says our methodological justification is unconvincing. Write a response." Assistant: "Let me use the paper-writer agent to draft a firm, evidence-grounded rebuttal." (Since the user needs a peer review response written, use the paper-writer agent.)

Use this agent when you need rigorous, adversarial peer review of any component of the manuscript — writing, figures, analyses, or the overall scientific claim — at the level of the target venue. This agent does not encourage; it identifies every weakness that would cause rejection and forces the team to address them before submission. Examples: - User: "Review the Introduction draft" Assistant: "Let me use the reviewer agent to evaluate the Introduction as a target-venue referee." (Since the user wants pre-submission critique of manuscript text, use the reviewer agent.) - User: "Is Figure 3 good enough for our target journal?" Assistant: "Let me use the reviewer agent to assess Figure 3 against target-venue standards." (Since the user needs a quality gate on a figure, use the reviewer agent.) - User: "Are there any holes in our methodology that a reviewer would attack?" Assistant: "Let me use the reviewer agent to identify methodological vulnerabilities." (Since the user needs adversarial stress-testing of the methods, use the reviewer agent.) - User: "We're about to submit — do a full manuscript review" Assistant: "Let me use the reviewer agent to conduct a complete pre-submission review across all dimensions." (Since the user needs a final quality gate before submission, use the reviewer agent.)

Three-step pipeline from a Keynote (.key) or PowerPoint (.pptx) deck to manuscript + outline. Func 1 exports vector PDFs per slide; func 2 crops them (vector preserved) and renders an outline-grade PNG from the same cropped artifact; func 3 copies PDFs into the LaTeX manuscript figures dir and PNGs into the outline.md figures dir, then inserts/updates image links after each result heading.

Drive one figure from design idea to manuscript-ready vector PDF + outline-ready PNG. Loops `@figure-descriptor` → `@analysis-implementer` → `@reviewer` against a single fig-id in `figures.json`, with the `cropfig` skill auto-invoked at the end of each implement phase to keep manuscript + outline artifacts in lockstep. The third Phase 2 engine — a 3-agent loop, more complex than `/iterate-revision`'s 2-agent loop, and the only Phase 2 engine that executes real code per iteration. Safe to re-run; safe to resume after BLOCKED or HALT.

Revise one manuscript section against the reviewer team until DONE, BLOCKED, or HALT. Loops `@paper-writer` ↔ `@reviewer` with a venue-specific reviewer brief, recording every iteration's issues + verdict to `reviews.json` and updating `paper.json.sections[name]` status / iter. The first OMCR engine — a worked example of how to compose `skills/orchestrate/phases/*` primitives into a domain-specific loop. Safe to re-run; safe to resume after BLOCKED or HALT.

Find, summarize, and verify N papers on a topic, then drop verified entries into `references.bib` and the literature summary CSV. Dispatches `@literature-curator` (one or many instances) over a CrossRef/OpenAlex candidate list, runs every survivor through the `verify-citation` skill, and writes the run summary to `citations.json.last_sweep`. Sequential by default; opt-in `--parallel N` (1 ≤ N ≤ 4) fan-out for speed. Safe to re-run; idempotent against duplicate DOIs (existing BibTeX entries are skipped, not double-added).

Scaffold a LaTeX manuscript directory for a research project — copy the bundled skeleton (main.tex + sections/* + figures/ + references.bib + .gitignore + README), optionally apply a journal-specific documentclass from templates/journal-registry.json, optionally clone an Overleaf project and cache the Git credential helper (token never persisted to tracked files), commit on the default branch, and ask before pushing. Invoked by /start-research phase 6, but also standalone-callable when adding a manuscript dir to an existing project later.

4 hooks across 3 events