[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-93293":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":8,"htmlUrl":8,"language":8,"languages":8,"totalLinesOfCode":8,"stars":9,"forks":10,"watchers":11,"openIssues":12,"contributorsCount":12,"subscribersCount":12,"size":12,"stars1d":12,"stars7d":13,"stars30d":13,"stars90d":12,"forks30d":12,"starsTrendScore":14,"compositeScore":15,"rankGlobal":8,"rankLanguage":8,"license":8,"archived":16,"fork":16,"defaultBranch":17,"hasWiki":18,"hasPages":16,"topics":19,"createdAt":8,"pushedAt":8,"updatedAt":20,"readmeContent":21,"aiSummary":22,"trendingCount":12,"starSnapshotCount":12,"syncStatus":23,"lastSyncTime":24,"discoverSource":25},93293,"Rebuttal-Skill","TobiasLee\u002FRebuttal-Skill","TobiasLee",null,317,12,114,0,203,9,67.34,false,"main",true,[],"2026-07-22 04:02:08","# Rebuttal Skill\n\nA structured AI skill for generating academic rebuttals to peer review. Designed for AI coding assistants (OpenCode, Claude Code, Gemini CLI) to produce rigorous, evidence-grounded author responses for venues like NeurIPS, ICML, ACL, and CVPR.\n\n## Features\n\n**Two-Stage Workflow**\n- **Stage 1 – Triage & Experiment Planning:** Parses reviews into atomic concerns, infers underlying reviewer intent, classifies severity (FATAL → MINOR), and produces a prioritized P0–P3 experiment plan — not an unranked wishlist.\n- **Stage 2 – Rebuttal Drafting:** Runs after author results arrive. Validates evidence against claimed concerns, drafts responses using a \"Direct Answer → Evidence → Revision\" structure, and produces a concrete manuscript revision list.\n\n**Rebuttal vs. Resubmission Gate (Stage 0)**\n- Assesses whether rebuttal is worthwhile (`PROMISING` \u002F `BORDERLINE` \u002F `LOW EXPECTED RETURN`).\n- For low-return cases, provides a detailed resubmission roadmap with rejection diagnosis, revision backlog, and next-submission experiment plan.\n\n**Response Patterns**\n- Templates for: correcting misunderstandings, acknowledging limitations, reporting experiments, novelty defense, missing baselines, statistical reliability, score-text mismatch, and more.\n\n**Five Output Modes**\n- `TRIAGE_AND_EXPERIMENT_PLAN` — before results are available\n- `RESULT_INTEGRATION` — after partial results arrive\n- `FULL_REBUTTAL` — complete rebuttal with opening summary and revision list\n- `RESUBMISSION_PLAN` — when rebuttal has low expected return\n- `QUALITY_REVIEW` — audit an existing rebuttal draft\n\n**Evidence Integrity**\n- Strictly prohibits fabricating values, presenting planned work as completed, or hiding negative results. Uses visible placeholders for missing information.\n\n## Usage\n\nThis is a skill definition, not a standalone application. Load it in any AI coding assistant that supports skills, then provide your paper's reviews:\n\n```\nAnalyze this paper's reviews before drafting a rebuttal.\n\nVenue and score scale: [venue, scale, borderline if known]\nDeadline and resources: [time remaining, compute, author bandwidth]\nAbstract: [abstract]\nReviews, scores, and confidences: [reviews]\n\nFirst assess rebuttal viability. Infer the underlying concern behind every\nreview comment and visibly flag uncertain interpretations. Then produce a\nprioritized P0-P3 experiment and analysis plan...\n```\n\n## Required Inputs\n\n| Stage 1 | Stage 2 (additional) |\n|---------|---------------------|\n| Paper abstract | Completed experiment results |\n| All review texts | Exact experimental settings |\n| Reviewer scores | Verified manuscript locations |\n| Venue & score scale | Changes authors are willing to make |\n| Deadline \u002F time remaining | Claims authors are willing to narrow |\n| Available compute & bandwidth | |\n","这是一个面向学术论文作者的AI rebuttal技能，专为生成严谨、证据驱动的同行评审回应而设计。核心功能包括两阶段工作流：第一阶段对评审意见进行分类 triage、推断审稿人意图并生成优先级实验计划；第二阶段基于实证结果，按‘直接回答→证据→修订’结构撰写回应，并输出具体修改清单。支持 rebuttal 可行性评估与拒稿后的重投路线图，强调证据完整性，禁止虚构数据。适用于 NeurIPS、ICML、ACL、CVPR 等顶会投稿作者在评审回复阶段辅助决策与文本生成。",2,"2026-07-15 02:30:11","CREATED_QUERY"]