Generate a structured paper outline from review conclusions and experiment results. Use when user says "写大纲", "paper outline", "plan the paper", "论文规划", or wants to create a paper plan before writing.

Paper Plan: From Review Conclusions to Paper Outline Generate a structured, section by section paper outline from: $ARGUMENTS Constants REVIEWER MODEL = gpt 6 astra — Model used via Codex MCP for outline review. Must be an OpenAI model. TARGET VENUE = ICLR — Default venue. User can override (e.g., /paper plan "topic" — venue: NeurIPS ). Supported: ICLR , NeurIPS , ICML , CVPR , ACL , AAAI , ACM , IEEE JOURNAL (IEEE Transactions / Letters), IEEE CONF (IEEE conferences). MAX PAGES — Page limit. For ML conferences: main body to Conclusion end (excluding references, appendix). ICLR=9, NeurIPS=9, ICML=8, AAAI=7 technical content pages plus references unless the current AAAI CFP says otherwise. For IEEE venues: references ARE included in page count. IEEE journal Transactions ≈ 12 14 pages total, Letters ≈ 4 5 pages total; IEEE conference ≈ 5 8 pages total (including references). Inputs The skill expects one or more of these in the project directory: 1. NARRATIVE REPORT.md or STORY.md — research narrative with claims and evidence 2. review stage/AUTO REVIEW.md — auto review loop conclusions (fall back to ./AUTO REVIEW.md if not found) 3. Experiment results — JSON files in figures/ , screen logs, tables 4. idea stage/IDEA REPORT.md — from idea discovery pipeline (if applicable) (fall back to ./IDEA REPORT.md if not found) 5. Compact files (if available): idea stage/IDEA CANDIDATES.md (fall back to ./IDEA CANDIDATES.md if not found) , findings.md , EXPERIMENT LOG.md — preferred over full files when present, saves context window If none exist, ask the user to describe the paper's contribution in 3 5 sentences. Orchestra Guided Writing Overlay Keep the existing insleep workflow and outputs, but use the shared references below to improve the quality of the story and outline. Read ../shared references/writing principles.md when framing the one sentence contribution, Abstract, Introduction, Related Work, or hero figure. Read ../shared references/venue checklists.md before freezing the outline for a specific venue. Only load these references when needed; do not paste their full contents into the working draft. Optional: Style reference ( — style ref: <source , opt in) Lets the user steer the structural layout of the outline (section ordering, subsection density, theorem environment density, figure budget, citation style) toward a reference paper. Default OFF — when the user does not pass — style ref , do nothing differently from before. Only when — style ref: <source appears in $ARGUMENTS , run the helper FIRST, before drafting the outline: Sources accepted: local TeX dir / file, local PDF, arXiv id ( 2501.12345 or arxiv:2501.12345 ), http(s) URL. Overleaf URLs and project IDs are rejected — clone via /overleaf sync setup <id first and pass the local clone path. Strict rules (full contract in tools/extract paper style.py docstring): Use style profile.md as structural guidance only when proposing the outline's section list, subsection counts, theorem density, figure budget. Never copy prose, claims, examples, section names verbatim, or terminology from anything reachable through the cache. The user's narrative is the only source of substance. Never pass — style ref (or the cache contents) to reviewer / auditor sub agents. Cross model review independence ( ../shared references/reviewer independence.md ) requires reviewers see only the artifact and the user's prompt. Gap Report ( GAP REPORT.md , auto emitted when style ref is on) When — style ref: succeeded AND any of figures/ , results/ , data/ , tables/ , sec/ , NARRATIVE REPORT.md , CLAIMS FROM RESULTS.md exists in the project, also emit a gap report before drafting the outline. The gap report maps the exemplar's section topology + density requirements (from style profile.md ) against the user's actual assets, surfacing structural slots where the user has no evidence to fill . It is the contract by which /paper write decides when to emit <! DATA NEEDED markers instead of fabricating content. Procedure: 1. Read $CACHE/style profile.md for exemplar's section list + per section feature counts (figures, theorems, tables, citations, sentences per section). 2. Inventory user assets: figures/ filenames, results/ evidence files, sec/ .tex existing prose, NARRATIVE REPORT.md , CLAIMS FROM RESULTS.md (if /result to claim ran), references.bib for citation density. 3. For each section slot the exemplar implies (ablation table, scaling experiment, failure case analysis, proof block, …), classify as covered / partial / missing . 4. Emit <output dir /GAP REPORT.md : Slot ID format: GAP <SECTION <FEATURE , all caps, stable across regenerations unless user assets change. Rules (hard): Do not infer, fill, or hallucinate evidence to "close" gaps. Missing is missing. Do not propose specific experiment commands to fill gaps — that is /experiment bridge 's job. Gap Report just surfaces deficits. Do not include exemplar prose / claim text / author names / quantitative figures from the exemplar. If style profile.md extraction failed or the user has no project assets, skip Gap Report (no error; just do not emit the file). The gap report is also subject to reviewer isolation — never passed to reviewer / auditor sub agents (same rule as style profile.md ). Original idea: @zhangpelf in [ 217](https://github.com/wanshuiyin/Auto claude code research in sleep/issues/217). Workflow Step 1: Extract Claims and Evidence First check for CLAIMS FROM RESULTS.md — if its first line is verdict: REVIEW UNAVAILABLE , treat the file as ABSENT for claim extraction (fall through to the narrative documents below) and then: under — assurance: submission ( shared references/assurance contract.md ; implied by — effort: max beast ) STOP — the claims were never adjudicated, rerun /result to claim first; under assurance: draft continue but tag every claim [unadjudicated] in the claims matrix. Otherwise, if it exists (generated by /result to claim at the end of Workflow 2), use it as the starting point for claims. This file contains validated claims already mapped to experiment evidence. Merge with any additional claims from the narrative documents below. If CLAIMS FROM RESULTS.md does not exist, extract claims from scratch: Read all available narrative documents and extract: 1. Core claims (3 5 main contributions) 2. One sentence contribution (the single sentence that best states what the paper contributes) 3. Evidence for each claim (which experiments, which metrics, which figures) 4. Known weaknesses (from reviewer feedback) 5. Suggested framing (from review conclusions) Build a Claims Evidence Matrix : Step 2: Determine Paper Type and Structure Based on TARGET VENUE and paper content, classify and select structure. Before committing to a structure, apply the narrative principle from ../shared references/writing principles.md : The paper should tell one coherent technical story. By the end of the Introduction, the outline should make the What , Why , and So What explicit. Front load the most important material: title, abstract, introduction, and hero figure. Reviewers often form a judgment before reading the full method. IMPORTANT : The section count is FLEXIBLE (5 8 sections). Choose what fits the content best. The templates below are starting points, not rigid constraints. Empirical/Diagnostic paper: Theory + Experiments paper: Theory papers often need 7 sections (splitting theory into estimation + optimization, or setup + analysis). The total page budget MUST sum to MAX PAGES. Theory papers should: Include proof sketch locations (not just theorem statements) Plan a comparison table of prior theoretical bounds vs. this paper's bounds Identify which proofs go in appendix vs. main body Method paper: Step 3: Section by Section Planning For each section, specify: Step 4: Figure Plan List every figure and table: CRITICAL for Figure 1 / Hero Figure : Describe in detail what the figure should contain, including: Which methods are being compared What the visual difference should demonstrate Caption draft that clearly states the comparison Why the figure helps a skim reader understand the paper before reading the full method Step 5: Citation Scaffolding For each section, list required citations: Citation rules (from claude scholar + Imbad0202/academic research skills): 1. NEVER generate BibTeX from memory — always verify via search or existing .bib files 2. Every citation must be verified: correct authors, year, venue 3. Flag any citation you're unsure about with [VERIFY] 4. Prefer published versions over arXiv preprints when available Step 6: Cross Review with REVIEWER MODEL Send the complete outline to GPT 6 Astra xhigh for feedback: Apply feedback before finalizing. Step 7: Output Save the final outline to PAPER PLAN.md in the project root: Key Rules Large file handling : If the Write tool fails due to file size, immediately retry using Bash ( cat << 'EOF' file ) to write in chunks. Do NOT ask the user for permission — just do it silently. Do NOT generate author information — leave author block as placeholder or anonymous Be honest about evidence gaps — mark claims as "needs experiment" rather than overclaiming Page budget is hard — if content exceeds MAX PAGES, suggest what to move to appendix MAX PAGES counting differs by venue — ML conferences: main body to Conclusion end, references/appendix NOT counted; AAAI main track is typically 7 technical content pages plus references. IEEE venues: references ARE counted toward the page limit. Venue specific norms — ML conferences (ICLR/NeurIPS/ICML) use natbib ( \citep / \citet ); IEEE venues use cite package ( \cite{} , numeric style) Claims Evidence Matrix is the backbone — every claim must map to evidence, every experiment must support a claim Front load the story — the outline should make the contribution clear in the title, abstract, introduction, and hero figure before the reader reaches the full method Figures need detailed descriptions — especially the hero figure, which must clearly specify comparisons and visual expectations Section count is flexible — 5 8 sections depending on paper type. Don't force content into a rigid 5 section template. Acknowledgements Outline methodology inspired by [Research Paper Writing Skills](https://github.com/Master cai/Research Paper Writing Skills) (claim evidence mapping), [claude scholar](https://github.com/Galaxy Dawn/claude scholar) (citation verification), and [Imbad0202/academic research skills](https://github.com/Imbad0202/academic research skills) (claim verification protocol). The writing framing overlay in this hybrid pack is adapted from Orchestra Research's paper writing guidance. Output Protocols Follow these shared protocols for all output files: [Output Versioning Protocol](../shared references/output versioning.md) — write timestamped file first, then copy to fixed name [Output Manifest Protocol](../shared references/output manifest.md) — log every output to MANIFEST.md [Output Language Protocol](../shared references/output language.md) — respect the project's language setting

paper-plan