agentleFS
jaccenjaccen/Awesome-Gaussian-Skills/skills/cg-paper-writing/SKILL.md

cg-paper-writing

Skill186 starsChanged 57 days agoApache-2.0View on GitHub

Academic paper writing for 3D vision, computer graphics, CAD, and 3D understanding. Covers NeRF, 3DGS, SLAM, point cloud, 3D shape, CAD modeling. Supports CVPR/ICCV/ECCV/SIGGRAPH venues and Chinese core journals (17 journal-specific format specs). Multi-agent adversarial review, citation integrity gates, style calibration, paper novelty/industry-value evaluation, pre-submission format audit, experiment claim verification. Use when: writing or revising a CG/3D vision paper, drafting abstract/intro/method/experiments, running adversarial review or citation integrity check, calibrating writing style to a venue, evaluating paper value, 写论文/写paper/CG论文/计算机学报投稿/软件学报投稿/图形学学报投稿/自动化学报投稿/中文核心期刊.

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 13 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈13

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill to improve research clarity, auditability and writing efficiency.

    line 46
  2. Distinctive:

    If the user does not specify, defaults are: section=all, venue=all.

    line 120
  3. Distinctive:

    Read these files from references/:

    line 125
  4. Distinctive:

    After loading the required fragments:

    line 203
  5. Distinctive:

    Follow section-specific templates from the loaded writing fragment

    line 205
  6. Distinctive:

    Use terminology and symbols from symbols-terminology.md

    line 207
  7. Distinctive:

    Chinese academic writing: Follow Chinese academic conventions (本文/我们/由此/表明)

    line 217
  8. Distinctive:

    No invented data: Never fabricate experimental results, method capabilities, or review outcomes.

    line 229
  9. Distinctive:

    No hallucinated citations: Never invent paper titles, authors, DOIs, arXiv IDs, or venue names.

    line 230
  10. Distinctive:

    No silent speculation: If you are uncertain about a technical detail, explicitly flag it with "[UNCERTAIN]" rather than presenting it as …

    line 231
  11. Distinctive:

    No method misattribution: Do not assign features, results, or mechanisms from one method to another.

    line 232
  12. Distinctive:

    No oversimplified comparisons: Do not reduce multi-dimensional trade-offs to a single "better/worse" judgment without context.

    line 233
  13. Distinctive:

    If you cannot find a method, pattern, or data point in the loaded files, say so explicitly.

    line 286

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

CG Paper Writing Engine (Router)

Evidence-first operating contract

Use this skill to improve research clarity, auditability and writing efficiency. Do not promise acceptance, predict acceptance probabilities from self-scores, or imply that simulated roles are actual journal reviewers. For conflicts among fragments, this contract and references/review-integrity.md take precedence over historical templates and scoring heuristics.

Start from the supplied artifacts: manuscript/version, research question, contribution type, target venue/year/track, available results and requested deliverable. Proceed with a bounded draft when inputs are incomplete; label missing evidence rather than inventing it. A focused edit does not require a full submission audit. Only certify submission readiness for checks actually completed.

For methods, experiments, novelty assessment and full-paper review, load domain-evidence-review.md. For Nature-family targeting, also load top-journal-strategy.md; resolve the exact journal and article type before applying its requirements.

Templates are optional rhetorical examples, not evidence or mandatory contribution counts. Public-paper critique and author-owned mock review differ from confidential assigned peer review: verify the venue's current AI/confidentiality rules before accepting review materials; do not process prohibited confidential submissions or generate prohibited official review content.

Architecture: Axis-driven static/dynamic router. Do NOT try to apply the writing logic from memory or from this router alone. Always load fragments from disk as described below.

Step 1 — Detect Request Axes

Analyze the user's request to determine axis values:

Axis: section

User Intentsection value
Writing or revising titletitle
Writing or revising abstractabstract
Writing or revising introductionintro
Writing or revising related workrelated-work
Writing or revising methodologymethod
Writing or revising experimentsexperiments
Writing contribution statementscontribution
Topic / venue selection for a 3DGS sub-fieldtopic
Looking up verified method landscape / baselines / datasets / metricsknowledge
Designing or drawing paper figures (architecture/data/comparison)figure
Responding to reviewer comments / rebuttal / revision letterrebuttal
Full paper or unspecified sectionall

Axis: task (lifecycle fusion, 2026-10-27)

User Intenttask valueFused from
Evaluate whether a research idea is worth pursuing / novelty check / score this ideaidea-evalidea-evaluator
Plan paper skeleton / logic chain / thinking template / structure planningskeletontech-paper-template
Search SCI journals/conferences for literature / build literature library / download PDFs / generate BibTeXliteraturesci-journal-literature + literature-manager
Design figures / draw architecture diagrams / plot experiment data / figure quality auditfigurefigure-designer + academic-plotting + drawio-academic
Design experiments / select datasets/baselines/metrics / plan ablation / debug training crashes (NaN/OOM/floaters)experiment3dgs-experiment-planner + 3dgs-training-debugger
Polish existing prose / remove AI tone / Chinese-to-English translation / calibrate claim strengthpolishpaper-polish + nature-polishing
Respond to reviewer comments / write rebuttal letter / cover letter / marked manuscriptrebuttalrebuttal-guidance + nature-response
No lifecycle task (pure section drafting)none—

Task axis takes precedence over section axis when both are detected. For example, "评估这个 idea" → task=idea-eval (ignore section); "画架构图" → task=figure (also sets section=figure); "回复审稿人" → task=rebuttal (also sets section=rebuttal). When task=none, fall back to section-driven routing.

Axis: venue

User Intentvenue value
Targeting CVPR/ICCV/ECCVcvpr
Targeting SIGGRAPH/EG/PGsiggraph
Targeting NeurIPS/AAAIneurips
Targeting TVCG/CGF/TOG/TPAMItvcg
Writing PhD thesis chapterthesis
Targeting Nature/Science sub-journals, Chemical Reviews, or other top-tier journals; submission strategy questions (投稿策略/顶刊/子刊/综述)top-journal
Targeting Chinese core journals (软件学报/计算机学报/计算机研究与发展/中文核心期刊) or writing in Chinesechinese-journal
Targeting 计算机学报 (Chinese Journal of Computers / CJC / 学报综述 / 学报投稿)cjc
Targeting 软件学报 (Journal of Software / JoS / 软件学报投稿)jos
Targeting 计算机辅助设计与图形学学报 (JCAD / 图形学学报投稿 / CAD与图形学)jcad
Targeting 自动化学报 (Acta Automatica Sinica / AAS)aas
Targeting 中国科学:信息科学 (Scientia Sinica Informationis / SCIS)scis
Targeting 计算机研究与发展 (JCRD / CRAD)crad
Targeting 中文信息学报 (JCIS / NLP学报)jcis
Targeting 中国图象图形学报 (JIG / 图象图形学报)cjig
Targeting 模式识别与人工智能 (PRAI / PR与AI)pra
Targeting 智能系统学报 (CAAI Transactions / 智能系统)caai
Targeting 虚拟现实与智能硬件 (VRIH / VR期刊)vrih
Targeting 计算机科学 (Computer Science / JSKX)jskx
Targeting 计算机应用 (Journal of Computer Applications / JCA)jsjy
Targeting 图学学报 (Journal of Graphics / TXB)txxb
Targeting 机器人 (Robot / 机器人学报)robot
Targeting 计算机应用研究 (Application Research of Computers / AROC)jsjyyj
Targeting 工医艺的可视计算 (VCIBA / 可视计算)vciba
Targeting CCF国际期刊 / IEEE/ACM Transactions / Springer / Elsevier 期刊投稿ccf-intl
Unspecified or multi-venueall

If the user does not specify, defaults are: section=all, venue=all.

Step 2 — Load Required Fragments

Always Load (every invocation)

Read these files from references/:

  • references/review-integrity.md — Shared evidence and scoped review rules
  • references/core-stance.md — Role, writing process, de-AI rules, citation fact-check, guardrails
  • references/symbols-terminology.md — Mathematical symbols, CG/CAD/3D terminology reference
  • references/writing-patterns-from-top-papers.md — Exemplar patterns from 3DGS/Mip-Splatting/D4RT (concrete writing techniques with direct quotes); load for style calibration on any section
  • references/core-paper-writing-paradigm.md — 领域通用论文架构模式(多源证据:10 篇 CVPR 2026 共性 + 3DGS/Mip-Splatting/D4RT 逐句拆解 + 四方向模板;两篇新增样本仅入证据附录):给“论证链六环 + 五类论文 + 八环节变体矩阵”的不变量与中英双轨模板;任何环节写作、结构规划、中文核心投稿、按范文校准文风时加载

On-Demand Load (by detected section)

sectionFragment(s) to Load
titlereferences/writing-abstract-intro.md
abstractreferences/writing-abstract-intro.md
introreferences/writing-abstract-intro.md
related-workreferences/writing-related-method.md + references/3dgs-paper-knowledge.md
methodreferences/writing-related-method.md
experimentsreferences/writing-experiments.md + references/3dgs-paper-knowledge.md
contributionreferences/writing-experiments.md
topicreferences/3dgs-domain-venues.md
knowledgereferences/3dgs-paper-knowledge.md
figurereferences/figure-design-protocol.md
rebuttalreferences/rebuttal-protocol.md
allAll writing fragments (abstract-intro, related-method, experiments) + references/3dgs-domain-venues.md

On-Demand Load (by detected task — lifecycle fusion)

When task axis is non-none, load the corresponding fragment FIRST, then supplement with section/venue fragments if the task also involves writing. Task fragments fuse 7 external skills into this skill for full-pipeline coverage (2026-10-27).

taskFragment(s) to LoadCovers lifecycle stage
idea-evalreferences/idea-evaluation.mdStage 1: 选题评估(五维+致命伤审计+范式转移探针)
skeletonreferences/paper-skeleton-protocol.mdStage 1: 论文骨架(思维模板+四点自洽)
literaturereferences/literature-search-protocol.mdStage 2: 文献检索(SCI/顶会,排除预印本)+ 下载 + BibTeX
figurereferences/figure-design-protocol.mdStage 4: 绘图(三核心图设计+matplotlib+drawio+QC)
experimentreferences/experiment-design-protocol.mdStage 5: 实验设计(数据集/基线/指标/消融)+ 训练调试(60种失败模式)
polishreferences/polishing-protocol.mdStage 6: 润色(投稿级+中译英+AI味去除+claim校准)
rebuttalreferences/rebuttal-protocol.mdStage 8: 返修(条件说服P(ZX,Y)+逐条指导+cover letter+marked manuscript)
none(fall back to section-driven routing)—

On-Demand Load (by detected venue)

venueFragment to Load
top-journalreferences/venue-formats.md + references/top-journal-strategy.md + references/ccf-international-journals.md
chinese-journalreferences/venue-formats.md + references/writing-chinese-journal.md + references/ccf-journal-catalog.md
cjcreferences/venue-formats.md + references/writing-chinese-journal.md + references/cjc-format.md + references/ccf-journal-catalog.md
josreferences/venue-formats.md + references/writing-chinese-journal.md + references/jos-format.md + references/ccf-journal-catalog.md
jcadreferences/venue-formats.md + references/writing-chinese-journal.md + references/jcad-format.md + references/ccf-journal-catalog.md
aasreferences/venue-formats.md + references/writing-chinese-journal.md + references/aas-format.md + references/ccf-journal-catalog.md
scisreferences/venue-formats.md + references/writing-chinese-journal.md + references/scis-format.md + references/ccf-journal-catalog.md
cradreferences/venue-formats.md + references/writing-chinese-journal.md + references/crad-format.md + references/ccf-journal-catalog.md
jcisreferences/venue-formats.md + references/writing-chinese-journal.md + references/jcis-format.md + references/ccf-journal-catalog.md
cjigreferences/venue-formats.md + references/writing-chinese-journal.md + references/cjig-format.md + references/ccf-journal-catalog.md
prareferences/venue-formats.md + references/writing-chinese-journal.md + references/pra-format.md + references/ccf-journal-catalog.md
caaireferences/venue-formats.md + references/writing-chinese-journal.md + references/caai-format.md + references/ccf-journal-catalog.md
vrihreferences/venue-formats.md + references/writing-chinese-journal.md + references/vrih-format.md + references/ccf-journal-catalog.md
jskxreferences/venue-formats.md + references/writing-chinese-journal.md + references/jskx-format.md + references/ccf-journal-catalog.md
jsjyreferences/venue-formats.md + references/writing-chinese-journal.md + references/jsjy-format.md + references/ccf-journal-catalog.md
txxbreferences/venue-formats.md + references/writing-chinese-journal.md + references/txxb-format.md + references/ccf-journal-catalog.md
robotreferences/venue-formats.md + references/writing-chinese-journal.md + references/robot-format.md + references/ccf-journal-catalog.md
jsjyyjreferences/venue-formats.md + references/writing-chinese-journal.md + references/jsjyyj-format.md + references/ccf-journal-catalog.md
vcibareferences/venue-formats.md + references/writing-chinese-journal.md + references/vciba-format.md + references/ccf-journal-catalog.md
any other non-all valuereferences/venue-formats.md
allreferences/venue-formats.md
ccf-intlreferences/venue-formats.md + references/ccf-international-journals.md

Reference Load (for review/integrity work)

  • references/review-integrity.md — Multi-agent review, Devil's Advocate Protocol(含矛盾消融结果消解模式), citation verification, integrity gates, style calibration, writing quality check, persistence
  • references/paper-evaluation-framework.md — 论文创新性与学术/产业价值评估框架(四维创新模型+TRL产业成熟度+10维评分表+审稿视角对照),用于投稿前自评、审稿、指导学生论文、评估他人工作
  • references/top-paper-patterns.md — 顶会论文共性模范范式(基于10篇CVPR 2026及前沿3DGS论文全文精读提炼):贡献结构配方(3-4条四件套+first声明+外部技术嫁接)、五条选题路径、四层实验证据体系、三种摘要模板、Intro四段式+根本问题句、图表引导策略、章节配比、顶会投稿自检清单、实测写作瑕疵避坑;写作/修改论文与校准结构时加载
  • references/core-paper-writing-paradigm.md — 领域通用论文架构模式(多源证据,2026-11-04):以“论证链六环(问题→缺口→洞察→机制→证据→边界)”为不变量,配五类论文(范式/诊断/系统/基准/分析/综述)与八环节(idea/题目/摘要/骨架/引言/方法/实验/结论)的变体矩阵,给中英双轨带槽位模板、通用检查清单与实例到规则的抽象方法;两篇新增样本 SeeLe 与 JIG 综述只入证据附录;写任何章节、规划结构、投中文核心、按范文校准文风时必读
  • references/reviewer-perspective.md — 审稿人视角拒稿与中稿规律(基于OpenReview 33条真实审稿意见编码):Champion效应与批评位置理论、拒稿理由Top10、录用要素Top10、可复述新颖性标准、致命伤vs可挽救伤判别与Rebuttal博弈(MoE-GS低分中稿解密)、跨venue转投策略、压缩方向八大坑、图表制作40条自查清单、投稿前防御清单D1-D15;投稿前自检、预判审稿攻击点、准备rebuttal、决定转投时加载
  • references/latex-source-hygiene.md — LaTeX 源码卫生与机械化验证(连续空行/环境与章节同行/跨文档引用写法/bib 键名一致性/S 编号对应/编译健康度/修复回归验证),配套一键核查脚本(维护者本地工具,未随技能分发,可按本文件方法论自写);投稿前格式核查与源码清理任务必读
  • references/markdown-draft-format-audit.md — Markdown 工作稿投稿前格式审计与 P0 机械修复(摘要词数/引用首现顺序/孤儿图表检测/彩色批注两类处置规则/占位符两步法引用重排/摘要瘦身/主文-补充切分标注/中英双稿独立同步与终检清单/章节重构后引用顺序复检/MD→LaTeX 转换边界 bibitem 核验);LaTeX 排版前的工作稿格式核查、批注清理与 P0/P2 修复任务必读
  • references/experiment-claim-verification.md — 实验声明溯源与服务器实验验证(声明-产物六步核查链/对比数字同维度核查与基线重跑/静默回退失效四联征/split fallthrough 陷阱与目录重命名法/h5→pkl 格式转换工作流/SSH-PowerShell-后台进程操作陷阱);评审中遇到「服务器规模验证」类声明溯源、撰写或核查对比实验表、为论文补做远程实验、排查「训练跑完 rc=0 但结果异常」时必读
  • references/3dgs-citation-audit.md — 3DGS 论文引文与声明核查(事实层):三层校验协议(ID 可达/名称一致/名称反查+人工裁定)、五类红线错误与本项目真实罚例(编号虚构、名↔号错配、venue 错、名/venue/描述全错、伪派生后缀)、实时引文核查脚本 scripts/audit_manuscript_citations.py 用法与 [RED FLAG]/[WARN] 判读、写作/投稿前/评审三套清单;写 related work 或实验时核对引用、投稿前自查引用真实性、评审他人稿件验证其引用与声称时必读(注意:任何事实结论必须回到 data/methods.json 与 .arxiv_cache.json 实时核查,不得凭本文件记忆作答)

Optimization: For a focused task (e.g., "write abstract"), load only core-stance + symbols-terminology + writing-abstract-intro + writing-patterns-from-top-papers + core-paper-writing-paradigm + venue-formats (if venue specified). For full paper work, load relevant writing and domain fragments; do not load unrelated venue specifications.

Step 3 — Execute Writing Task

After loading the required fragments:

  1. Follow section-specific templates from the loaded writing fragment
  2. Apply venue-specific formatting from venue-formats.md
  3. Use terminology and symbols from symbols-terminology.md
  4. Observe all guardrails from core-stance.md (no fabrication, de-AI rules, citation fact-check)
  5. For review/integrity work, follow protocols from review-integrity.md

Rules

  1. Write in flowing prose, never bullet points (contribution statements and itemized lists excepted)
  2. Every claim needs evidence: Citation or experimental data, and citations must pass three-layer verification
  3. Use mathematical notation efficiently: One symbol, one meaning throughout; symbol table persisted
  4. Match the venue's tone: CVPR more concise; SIGGRAPH more narrative; if style sample provided, strictly calibrate
  5. Chinese academic writing: Follow Chinese academic conventions (本文/我们/由此/表明)
  6. Never fabricate data: Mark missing data as <!-- DATA_NEEDED: <description> -->
  7. Integrity status: Complete applicable checks before calling an artifact submission-ready; drafts may retain explicit unresolved items.
  8. Evidence-led review: Revise or withdraw findings when better evidence appears. Do not enforce concession quotas or fixed debate rounds.
  9. Claim-citation alignment: Each claim must be traceable to supporting citation, and citation must actually support the claim
  10. Writing context persistence: Maintain symbol, citation, review state consistency across sessions
  11. Comparison dimension consistency: In comparison experiments, all compared methods (including baselines) must be evaluated on the same dataset — same version, resolution, and split — under the same evaluation protocol. When a baseline number's caliber cannot be verified, rerun the comparison under unified settings; never reuse, silently convert, or approximate a mismatched number, and never mix numbers from different dimensions in one table

Red Lines

The following are categorical prohibitions. Violating any of these invalidates the output:

  • No invented data: Never fabricate experimental results, method capabilities, or review outcomes. If a value is not found in the loaded files, write "data not available" or "N/A".
  • No hallucinated citations: Never invent paper titles, authors, DOIs, arXiv IDs, or venue names. Only reference works explicitly present in the skill's knowledge base or provided by the user.
  • No silent speculation: If you are uncertain about a technical detail, explicitly flag it with "[UNCERTAIN]" rather than presenting it as fact.
  • No method misattribution: Do not assign features, results, or mechanisms from one method to another. Each method's data is specific to that method.
  • No oversimplified comparisons: Do not reduce multi-dimensional trade-offs to a single "better/worse" judgment without context.
  • No cross-dimension comparisons: Numbers evaluated on different datasets, resolutions, splits, or evaluation protocols must never share one comparison table. When calibers mismatch, rerun under unified settings (writing-experiments.md → Comparison Fairness Protocol; experiment-claim-verification.md §6) — copying, silently converting, or approximating a mismatched number is a violation.

Extension Protocol — 新增中文期刊专项(Add a New Journal Venue)

当用户要求「搜索某期刊投稿要求,形成(这样的)能力/能力补充」时,按以下七步扩展本技能(已验证三轮:计算机学报 cjc / 软件学报 jos / 计算机辅助设计与图形学学报 jcad):

  1. 采集官方一手来源(按优先级,禁止凭记忆猜测格式):
    • 官方模板 .doc/.docx → Word COM 提取全文(New-Object -ComObject Word.Application,取 $doc.Content.Text);
    • 官方投稿须知/修改稿要求 PDF → PyMuPDF(fitz/pymupdf)提取文本;
    • 官网投稿指南页 → webfetch;若为 JavaScript 动态渲染(如 jcad.cn)只返回导航框架,降级为 online_search 检索 + 权威二手来源(主办单位介绍页、投稿经验帖)交叉验证,并在产出文件中标注来源与可信度。
  2. 提炼硬性规范:期刊基本信息(ISSN/CN/收录/分级/审稿制度/周期/费用)、投稿要求、结构要求(研究论文 vs 综述)、首页要素与字号、摘要要素与字数、关键词数量、参考文献格式、署名规范、常见退稿原因、投稿前 CheckList,并附与已收录期刊的差异对比表。
  3. 创建 references/<venue>-format.md:开头注明官方来源(模板文件名/网址/规定日期);无法核实的细节标注「以官网下载区模板/编辑部通知为准」。
  4. 登记路由:SKILL.md 的 venue 检测表、venue 加载表、when_to_use 触发词各加一行;源仓库副本的 manifest.yaml 同步 venue values 与 on_demand 路由。
  5. 衔接通用指南:writing-chinese-journal.md 加提示——目标为该刊时必须额外加载 <venue>-format.md,与通用条目冲突时以专项文件(官方来源)为准。
  6. 双副本同步与发布纪律:本技能为双副本安装(安装目录 ~/.workbuddy/skills/cg-paper-writing/ + 源仓库 Awesome-Gaussian-Skills/skills/cg-paper-writing/),内容改动须在同一会话内同步两处(两副本结构一致:SKILL.md / manifest.yaml / references/,可整文件覆盖);发布包 dist/cg-paper-writing.zip 由源副本重新打包生成。发布前必须重跑隐私扫描(scripts/scan_skill_privacy.py),确认不含未发表稿件相关材料(稿件项目名、方法名、实验数字、数据集名、AIGC 设备标识码等)。
  7. 路由一致性校验:批量编辑后须逐项核验 SKILL.md venue 轴、On-Demand 路由表与 manifest.yaml 的 axes.venue.values、routing.on_demand.venue 一一对应(曾因 multiedit 回退路由表丢失 14 个 venue 路由,验证时才发现)。

Cross-Skill Routing

This skill covers the full paper lifecycle (8 stages) via the fused task axis (2026-10-27). Most lifecycle capabilities are now in-skill; the following remain external:

Already fused into this skill (task axis, no longer need external skill):

  • idea-evaluator → task=idea-eval (references/idea-evaluation.md)
  • tech-paper-template → task=skeleton (references/paper-skeleton-protocol.md)
  • sci-journal-literature + literature-manager → task=literature (references/literature-search-protocol.md)
  • figure-designer + academic-plotting + drawio-academic → task=figure (references/figure-design-protocol.md)
  • 3dgs-experiment-planner + 3dgs-training-debugger → task=experiment (references/experiment-design-protocol.md)
  • paper-polish + nature-polishing → task=polish (references/polishing-protocol.md)
  • rebuttal-guidance + nature-response → task=rebuttal (references/rebuttal-protocol.md)

Still external (not fused, route out when needed):

  • Method comparison → 3dgs-method-compare (for positioning and related work tables,挂接方法库)
  • Paper reading/analysis → 3dgs-paper-reader / 3dgs-visualizer (for deep reading and method-library visualizations)
  • Code review → 3dgs-code-reviewer (for pre-training static code analysis, 104 bug patterns)
  • Autonomous experiment orchestration → autoresearch (for two-loop "sleep research" — runtime capability, not a knowledge fragment)
  • Full-pipeline orchestration → academic-pipeline (for 10-stage hands-off orchestration — runtime orchestrator, not a knowledge fragment)
  • Bilingual delivery → bilingual-paper-sync (for Word 4-piece EN/CN docx+md — format conversion, not writing knowledge)

Related Skills

  • 3dgs-paper-reader — Deep reading and analysis of 3DGS papers (use for understanding related work)
  • 3dgs-method-compare — Method comparison (use for positioning and related work tables)
  • 3dgs-code-reviewer — Code review (use for pre-training implementation verification)
  • 3dgs-visualizer — Method-library visualization (radar/timeline/comparison, for paper figures with method-db)
  • autoresearch — Autonomous experiment orchestration (two-loop, for hands-off experiment runs)
  • academic-pipeline — Full-pipeline orchestrator (10-stage, for hands-off write→review→revise)
  • bilingual-paper-sync — Bilingual Word delivery (EN/CN docx+md 4-piece)

Guardrail: Do Not Apply From Memory

Do NOT try to apply the logic, method data, bug patterns, or technical details described in this skill from memory. Always read the SKILL.md and referenced files from disk before producing any output. The knowledge base is updated frequently; stale memory may produce outdated, inaccurate, or fabricated results.

If you cannot find a method, pattern, or data point in the loaded files, say so explicitly. Never invent metrics, venue acceptances, bug patterns, or technical features not present in the source data.

If you like it, please star this repo https://github.com/jaccen/Awesome-Gaussian-Skills

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.