--- spdx-license: AGPL-3.0-or-later user-invocable: true description: "Academic paper development — from first principles through submission." --- !`bash ~/.claude/hooks/sweetclaude/record-event.sh skill_invoked "sweetclaude:documents-academic-research" 2>/dev/null || true` !`cat .sweetclaude/state/session-state.yaml 2>/dev/null || echo "STATE_NOT_FOUND"` STOP. Before executing this skill, check: does .sweetclaude/state/phase.yaml exist in the project directory? If NO, do not proceed. Instead say: "This project is not configured for SweetClaude. Let me run the pre-flight check." Then invoke the sweetclaude master skill (Skill tool, skill: "sweetclaude:master") and run its pre-flight. Return here only after the pre-flight passes. # Academic Paper Development Six-phase pipeline from first principles to submission. Each phase has quality gates that must pass before advancing. The user controls pacing. **Model preference:** Phases 0 and 3 (argumentation) benefit from the highest-performance model available. Flag this at session start. --- ## Phase 0: First Principles **Purpose:** Establish the intellectual foundation before any literature or venue considerations. ### Process 1. **Core thesis.** Ask: "What is the central claim of this paper?" Probe until it is a single, clear statement that could be true or false. 2. **Key concepts.** Identify the 3-5 concepts the paper depends on. For each: what does it mean in this context, and what does the reader need to understand before encountering it? 3. **Novelty.** Ask: "What does not exist in the literature yet that this paper provides?" Push for specificity. "A new approach to X" is not enough. What specifically is new? 4. **Strongest objections.** Propose 2-3 objections a skeptical reviewer would raise. Discuss with the user: which are addressable, which are acknowledged limitations? 5. **Scope boundary.** What is this paper NOT about? What related topics does it deliberately exclude? ### Integration If canonical or canonical-draft documents exist in `strategy/academic-research/` for this topic (from reconciliation), read them as starting context. Do not start from zero if the corpus has material. If `strategy/narrative-arc/` exists, check: does this paper serve an objective in the arc? What claims does it need to support? ### Quality Gate 0 - [ ] Core thesis is a single, falsifiable statement - [ ] 3-5 key concepts identified and defined - [ ] Novelty is specific (not "a new approach") - [ ] At least 2 objections identified with response strategy - [ ] Scope boundary explicit (at least 2 things this paper is NOT about) --- ## Phase 1: Literature & Positioning **Purpose:** Understand what exists, identify gaps, and position this paper against prior work. ### Process 1. **Literature search — three rounds** (adapted from academic-paper-skills): - Round 1: Direct keyword search — 20-30 papers on the core topic - Round 2: Expanded search — 10-15 papers from adjacent fields, cited-by chains - Round 3: Foundational works — 5-10 seminal papers the field builds on - Use RAG index (`strategy/rag-index/`) and web search in combination 2. **Gap identification.** For each gap claimed: - State the gap clearly - Cite 3-5 papers that approach but don't fill the gap - Explain why existing work falls short - Connect back to the Phase 0 novelty claim 3. **Positioning against prior work.** For the 5-8 most relevant papers: - How does this paper differ? - What does this paper do that the prior work cannot? - Where does it build on vs. depart from prior work? 4. **SWOT analysis on the thesis.** - Strengths: what makes the argument compelling? - Weaknesses: what are the acknowledged limitations? - Opportunities: what does this open up for future work? - Threats: what could undermine the argument? (reproducibility, scope, competing work) ### Quality Gate 1 - [ ] 35+ papers reviewed across three rounds - [ ] At least 3 gaps identified, each with 3+ citations as evidence - [ ] 5-8 prior works positioned against with specific differentiators - [ ] SWOT completed with actionable items (weaknesses addressed in paper or flagged as limitations) --- ## Phase 2: Structure & Venue **Purpose:** Select target venue, learn its norms, and build a quality-gated outline. ### Process 1. **Venue selection.** Based on thesis, positioning, and intended audience, propose 2-3 target venues (journals, conferences, workshops, preprint platforms). For each: - Scope fit (does this paper's topic match?) - Audience fit (who reads this venue?) - Impact and prestige - Timeline (submission deadlines, review cycles) - Recommend one with rationale. User decides. 2. **Writing norms analysis.** Read 3-5 recent papers from the target venue: - Typical length and section structure - Citation style and density - Tone and formality level - What makes papers in this venue successful? 3. **Outline construction.** Build a 3-level hierarchical outline: - Level 1: Major sections (Introduction, Related Work, [Core sections], Discussion, Conclusion) - Level 2: Subsections with purpose statements - Level 3: Key points, arguments, and evidence per subsection - Word count targets per section (informed by venue norms) 4. **Reviewer-perspective assessment.** Score the outline on 7 dimensions (adapted from academic-paper-skills): | Dimension | What It Measures | Score | |---|---|---| | Clarity | Is the argument structure clear from the outline? | /5 | | Completeness | Are all necessary sections and sub-arguments present? | /5 | | Literature support | Does the outline reference sufficient evidence? | /5 | | Methodological clarity | Is the approach/methodology evident? | /5 | | Originality | Does the outline convey what's new? | /5 | | Organization | Does the flow make logical sense? | /5 | | Venue fit | Does the structure match the target venue's norms? | /5 | **Threshold: ≥28/35.** If below, revise outline before proceeding. ### Quality Gate 2 - [ ] Target venue selected with rationale - [ ] 3-5 sample papers analyzed for norms - [ ] 3-level outline complete with word count targets - [ ] Reviewer assessment ≥28/35 --- ## Phase 3: Modular Drafting **Purpose:** Write the paper section by section, with quality checkpoints after each. **Model preference:** Use the highest-performance model available. Argumentation quality is the bottleneck. ### Writing Sequence Write sections in this order (not the order they appear in the paper): 1. **Core argument sections** — the novel contribution (write these first while the argument is freshest) 2. **Related work** — position against literature from Phase 1 3. **Methodology/approach** — if applicable 4. **Introduction** — frame the paper now that you know what it says 5. **Discussion** — implications, limitations, future work 6. **Conclusion** — summarize contributions and significance 7. **Abstract** — write last (250-300 words, captures the whole paper) ### Per-Section Quality Checkpoint After each section, assess on 5 dimensions (adapted from academic-paper-skills): | Dimension | What It Measures | Score | |---|---|---| | Argument | Is the reasoning sound and well-supported? | /4 | | Citations | Are claims backed by appropriate references? | /4 | | Clarity | Is the writing clear and precise? | /4 | | Structure | Does the section flow logically? | /4 | | Consistency | Does it align with other sections already written? | /4 | **Threshold: ≥16/20.** If below, revise before moving to next section. ### Cross-Reference Check After all sections are drafted: - Verify terminology is consistent throughout - Verify forward/backward references are accurate - Verify no section contradicts another - Verify the abstract accurately reflects the final paper ### Quality Gate 3 - [ ] All sections drafted - [ ] Each section scored ≥16/20 - [ ] Cross-reference check passed - [ ] Abstract written and matches paper content --- ## Phase 4: Review & Revision **Purpose:** Simulate peer review, incorporate feedback, iterate. ### Process 1. **7-dimension reviewer simulation.** Score the complete manuscript: | Dimension | What It Measures | Score | |---|---|---| | Originality | Does it make a genuine contribution? | /10 | | Significance | Does the contribution matter to the field? | /10 | | Soundness | Is the reasoning valid and methodology appropriate? | /10 | | Clarity | Is the paper well-written and understandable? | /10 | | Literature | Is prior work adequately covered and positioned against? | /10 | | Completeness | Are there missing arguments, sections, or evidence? | /10 | | Venue fit | Is this paper appropriate for the target venue? | /10 | **Threshold: ≥56/70.** If below, identify the weakest dimensions and revise. 2. **Expert caucus review.** Invoke the `caucus` skill for multi-perspective review. Request perspectives relevant to the paper's domain (e.g., for an AI safety paper: security researcher, ML researcher, formal methods expert). 3. **Revision loop.** - Compile findings from reviewer simulation + caucus - Prioritize: critical issues → important → minor - Revise the manuscript - Re-run reviewer simulation on revised sections - Repeat until ≥56/70 with no critical findings 4. **Final assembly.** Ensure all sections are coherent as a whole. Read the paper end-to-end one final time. ### Quality Gate 4 - [ ] Reviewer simulation ≥56/70 - [ ] Expert caucus completed, findings addressed - [ ] No critical findings remaining - [ ] End-to-end read completed --- ## Phase 5: Submission **Purpose:** Format, finalize, and submit. ### Process 1. **Formatting.** Apply target venue's formatting requirements: - Template (LaTeX, Word, markdown) - Citation style (APA, IEEE, ACM, etc.) - Page/word limits - Figure/table formatting 2. **Abstract and metadata.** Finalize: - Abstract (250-300 words, last revision) - Keywords (5-8, informed by venue's taxonomy) - Author information and affiliations - Acknowledgments 3. **Submission checklist.** - [ ] Paper meets venue's formatting requirements - [ ] Page/word count within limits - [ ] All references are complete and properly formatted - [ ] Supplementary materials prepared (if applicable) - [ ] Cover letter drafted (if required) - [ ] All co-authors have reviewed and approved 4. **Submit.** Follow the venue's submission process. 5. **Post-submission.** Create a tracking document in `strategy/academic-research/`: ``` strategy/academic-research/{paper-slug}-submission-tracker.md ``` Contents: submission date, venue, expected review timeline, reviewer feedback (when received), revision plan. ### Quality Gate 5 - [ ] Formatted per venue requirements - [ ] Abstract finalized - [ ] Submission checklist complete - [ ] Tracking document created --- ## Narrative Arc Integration This skill reads the narrative arc but does not write to it. **Before Phase 0:** Check `strategy/narrative-arc/` for: - Does this paper serve an objective in the arc? - What claims does the arc need this paper to support? - What proof points should this paper strengthen? **After Phase 4:** Report what the paper contributes to the arc. The user decides whether and how to update the arc via the narrative-arc skill. ## Reconciliation Integration If `strategy/academic-research/` contains files from reconciliation (versioned source materials, canonical-drafts, or canonical docs), read them as starting context for Phase 0. Don't start from zero if the corpus has material on this topic.