jiancui-research
UserDo research in documents, not code — Spec-Driven Development for research papers, as slash commands for your AI coding agent (Claude Code, Codex, Copilot, OMP)
Categories
Indexed Skills (15)
research.ae
Prepare an artifact-evaluation submission (reproducibility checklist, artifact README, badge plan, permanent archival link).
research.analyze
Cross-artifact consistency + review-readiness audit AND the sync checker across plan, tasks, code, evidence, and manuscript. Read-only; outputs a prioritized gap report to .research/analyze-report.md.
research.constitution
Establish or update the research constitution (quality principles + writing voice + venue norms) at .research/memory/constitution.md
research.feasibility
De-risk the core idea with one small, cheap probe before committing to the full study; write .research/feasibility.md with a GO / NO-GO / PIVOT verdict.
research.implement
Work the tasks.md queue across Setup, Build, Eval, explicit user-led manuscript work, and Polish; keep claims and task status current.
research.init
Copy the bundled research-kit templates into this paper repo's .research/templates/ so the other commands can load them.
research.plan
Turn the proposal + feasibility into the study's technical design at .research/plan.md - architecture, evaluation design, key decisions, project layout. Stable; tasks derive from it.
research.proposal
Pipeline entry. Turn a raw research idea into proposal.md, a readable 1-3 page argument (falsifiable thesis, argued gap, pre-committed validation plan, venue + paper-type).
research.rebuttal
Draft a prioritized, evidence-backed rebuttal to reviewer comments into .research/rebuttal/, fitted to the venue word limit.
research.relatedwork
Survey prior work into .research/related-work.md AND sharpen the proposal's gap and positioning.
research.review
Self-review panel. Simulate reviewers reading ONLY the submitted paper, score it, and list specific findings each with a suggested fix command. Writes .research/review/round-N.md only; never edits the paper or any other artifact.
research.style
Build and keep the writing-style file the writing commands read - distilled from papers you admire, the instructions you give, and the edits you make.
research.tasks
Derive the single work queue .research/tasks.md from plan.md - Setup/Build/Eval/Paper/Polish sections, T-ids, claim links; re-runs refine and preserve checkbox states.
research.texreview
Open texreview, a local web UI to review the compiled paper - LaTeX source left, PDF right, SyncTeX click-to-source, comments on selections in either pane (button, `⌘⇧M`, or right-click) (requires uv + a TeX install). Comments are sidecar JSON in the paper repo's .texreview/ that any agent can read.
research.write
Write or revise one manuscript section - reads the whole paper and states its argument before any prose, then outlines, revises, critiques, or drafts.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.