clinpub

Phase 2: Adaptive statistical analysis. Claude diagnoses data structure (groups, timepoints, outcome types), proposes a customized analysis plan, discusses with user, and executes in dependency order. Each method outputs figure + table + README with publication-grade standards.

Phase 1: Data preparation and exploratory data analysis. Clean raw data, handle missing values, detect outliers, create derived variables, generate data quality report, and produce cleaned.csv.

Topic mining from clinical data tables (CSV/XLSX). Analyze variable structure, distribution patterns, missing data, and correlations; combine with PubMed literature search to identify research gaps; generate 3-5 structured candidate paper topics with feasibility scores. No statistical analysis or manuscript writing involved.

Read workspace state and auto-route to the appropriate clinpub command. With natural language input (e.g., '/clinpub:do 我想改清洁逻辑'), routes by intent. With no arguments, shows current state summary and suggests next commands.

Phase 0: Initialize or import a clinical research project. Detect existing artifacts and import into clinpub structure, or start fresh. Discuss study design, variables, analysis methods with user; generate project_config.yml, directory structure, and .clinpub/ artifacts.

Phase milestone management. Review completed phase deliverables, verify success criteria, record decisions, and gate progression to next phase. Generates MILESTONE.md and updates ROADMAP.md.

Modify completed analysis outputs. Clarifies modification scope (figure style, statistical method, variables), executes changes, verifies outputs, and records history in PLAN.md. Can be invoked from any phase. Triggers: modify figures, change analysis method, adjust chart style, replace variables.

Auto-advance to the next Phase or Wave. Verifies current step completion, auto-detects granularity (Wave vs Phase), updates STATE.md and ROADMAP.md, generates MILESTONE.md for phase transitions, and outputs clear prompt with next steps.

Clinical data analysis and publication pipeline command reference. Lists available phase commands and overview. Each phase must be invoked individually to ensure rigor. Targets SCI Q1/Q2 journals.

Phase 4: Simulated peer review and revision. Generate mock peer review, confirm revision items with user, revise manuscript, generate response letter. Loops until satisfactory.

Phase 3: IMRAD manuscript writing. Supports two modes: (1) batch mode (一鍵成稿) - bulk reference search then bulk writing with single final review; (2) sequential mode - per-section search, write, and user review. Shared reference library for deduplication. Placeholder-based cross-references. Final concatenation produces manuscript.md with YAML frontmatter.