
30 results

Synthesize all section drafts into a structured 250-word abstract (Background, Objective, Method, Results, Conclusion). Triggers when user says "/aw-abstract", "write abstract", "摘要", or called at end of Phase 2 before aw-finalize. Reads all section drafts, research-brief.json, and methodology.md to produce IMRAD-structured abstract.

GSDAW Discuss Agent #1 — Research Brief confirmation checkpoint. Triggers automatically after aw-questioner completes. Reads .planning/research-brief.json and asks user to confirm the research brief before proceeding to Research Agent + Methodology Agent. Shows summary of all 5 categories and offers confirm/supplement/restart choices.

GSDAW Discuss Agent #2 — Research vs Methodology consistency check. Triggers automatically after both aw-research and aw-methodology complete. Compares Literature Summary vs Methodology Design for consistency: - Does literature confirm the gap the methodology addresses? - Do the baselines in methodology appear in literature? - Are the datasets used realistic? Offers resolution options if conflicts found.

GSDAW Wave Execution Orchestrator — reads wave plan, spawns section-writing subagents in worktree isolation, runs quality gates, merges paragraph files into section files. Triggers when user runs /aw-execute or "开始执行" with a ready ROADMAP. For each wave: spawn subagents in parallel (one per independent task), wait for completion, quality gate checkpoint, merge paragraph files into section file via \input{}.

Generate figures from methodology.md figure list. Pipeline diagrams to PlantUML, results plots to matplotlib. Auto-inserts \input{figures/fig-name} at correct locations in section files. Triggers when user says "/aw-figure", "generate figures", "制图", or after Phase 2 sections are written.

GSDAW Methodology Agent — technical methodology design. Triggers when user approves Research Brief (after aw-discuss-1), or runs /aw-methodology. Reads research-brief.json to understand the problem and approach, then designs the full technical methodology: pipeline architecture, experiment plan, risk assessment. Outputs to .planning/methodology.md. Runs parallel with aw-research.

GSDAW Master Orchestrator — coordinates the full Phase 1 initialization flow. Triggers when user runs /aw-init or "开始写论文" with no active project. Coordinates: aw-questioner → aw-discuss-1 → [aw-research + aw-methodology parallel] → aw-discuss-2 → aw-planner → aw-discuss-3 → ready for /aw-execute. Supports --quick flag to skip Discuss checkpoints.

GSDAW Planning Agent — creates coarse-grained writing roadmap. Triggers after aw-discuss-2 confirms consistency, or runs /aw-plan. Reads research-brief.json, literature.md, and methodology.md. Outputs ROADMAP.md (by-chapter phases) and initial STATE.md. Each chapter = one phase with tasks and success criteria.

GSDAW User Questioner Agent — entry point for new academic paper projects. Triggers when user says "写论文", "开始写作论文", "new paper project", "/aw-init", or wants to start a new research paper. Deep questioning across 5 categories (Research Question, Approach, Methodology, Constraints, Materials) to generate a Research Brief. Auto-detects and offers to reuse prior brief. Supports /aw-init --quick for skipping Discuss checkpoints.

GSDAW Research Agent — literature search and analysis. Triggers when user approves Research Brief (after aw-discuss-1), or runs /aw-research. Reads research-brief.json to understand the research question, then searches the user's Zotero library (local SQLite, then HTTP API fallback, then PDF folder). Analyzes top 20 papers, generates literature matrix, identifies research gaps. Output: .planning/literature.md.

GSDAW Section Quality Review — runs automated checks and adversarial critique after each wave completes. Triggers automatically after each wave quality gate (called by aw-execute), or manually when user runs /aw-review. Automated checks: LaTeX compiles, word count, citations resolve, no TODO/FIXME. Presents results, asks user: "继续 / 修改 / 暂停".

GSDAW Wave Planner Agent — groups ROADMAP tasks into execution waves. Triggers when user runs /aw-wave-planner or approves a ROADMAP (after aw-planner). Reads ROADMAP.md, literature.md, methodology.md, and research-brief.json. Builds a dependency graph, performs topological sort, groups tasks into waves. Checks intra-wave file conflicts (tasks modifying same section file are sequential within a wave). Output: .planning/wave-plan.md.

Write the conclusion section of an academic paper. Triggers when user requests "conclusion", "结论", "/write conclusion", or when the wave executor assigns a conclusion writing task. Reads all previous section files (introduction, related-work, methodology, experiment, results, discussion) to synthesize contributions, key findings, and future work. Produces three paragraph files: 7-1-summary.tex, 7-2-findings.tex, 7-3-future-work.tex, plus abstract.tex last.

GSDAW Discussion Section Writer — writes Phase 6 (Discussion) of an academic paper. Triggers when user approves Results via quality gate after aw-write-results, or explicitly runs /aw-write-discussion. Reads results.tex, literature.md, and methodology.md (risks). Outputs independent paragraph .tex files to sections/discussion/ for wave aggregation.

GSDAW Experiment Section Writer — generates Phase 4 (Experiment) paragraph files. Triggered by /aw-write-experiment or by aw-execute during wave execution. Reads .planning/methodology.md (datasets, baselines, metrics, ablation sections) to generate experiment section paragraphs in Elsevier LaTeX format. Outputs independent paragraph .tex files to sections/experiment/ directory.

Write introduction section paragraphs for academic papers using wave-based execution. Triggers when /aw-execute assigns a Wave 1 or Wave 2 task for the introduction section (1.1, 1.2, 1.3, 1.4). Reads research-brief.json, literature.md, and methodology.md to produce independent paragraph .tex files. Each task writes one paragraph file following Elsevier LaTeX format.

GSDAW Methodology Section Writer — generates Phase 3 (Methodology) paragraph files. Triggered by /aw-write-methodology or by aw-execute during wave execution. Reads .planning/methodology.md and .planning/research-brief.json to generate methodology section paragraphs in Elsevier LaTeX format. Outputs independent paragraph .tex files to sections/methodology/ directory.

Write related work section paragraphs for academic papers using wave-based execution. Triggers when /aw-execute assigns a wave task for the related work section (2.1, 2.2, 2.3, 2.4). Reads literature.md categorized by theme to produce independent paragraph .tex files organized by category and method. Each task writes one paragraph file with proper citations following Elsevier LaTeX format.

GSDAW Results Section Writer — writes Phase 5 (Results) of an academic paper. Triggers when user approves Methodology via Discuss #3 and runs /aw-execute, or explicitly runs /aw-write-results. Reads methodology.md for metrics/baselines and experimental data/notes. Outputs independent paragraph .tex files to sections/results/ for wave aggregation.

Write academic paper sections in LaTeX with Elsevier format compliance. Triggers when user explicitly requests section writing via /write command, says "write section", "撰写章节", or asks to produce LaTeX content for a specific paper section. Does NOT activate passively on .tex file opens. Produces properly formatted LaTeX content with correct markup, citations, and academic voice.

Use when the user starts a new paper, asks to create a new paper while already on another paper branch, mentions creating a dedicated paper branch, or inputs "/paper-branch". This skill must reuse git-flow-branch-creator and create a title-based branch name so each paper is isolated on its own branch.