Skills Plugins MCP Prompt Model 博客 我的中心
内容创作 #design #writing #ai

paper-novelty-design-v1

Design, audit, and rewrite academic paper novelty claims into defensible contributions. Use when Codex needs to help with 论文创新点设计, novelty evaluation, contribution drafting, Problem-Method-Insight rewriting, 可证伪性/falsifiability checks, reviewer pressure tests, pseudo-innovation diagnosis, or innovation evolution logs for papers, theses, proposals, and technical reports.

DeepseekModel 官方收录技能 质量 优秀 · 78 v1.0.0

获取

https://deepseekmodel.com/api/download.php?id=lavineleo-paper-novelty-design-skill-md&format=skill
下载 .skill 标准格式,含 system_prompt 与 model_config,导入任意 Agent 框架即可使用
.skill 文件中 system_prompt 字段的实际内容。
name paper-novelty-design-v1 description Design, audit, and rewrite academic paper novelty claims into defensible contributions. Use when Codex needs to help with 论文创新点设计, novelty evaluation, contribution drafting, Problem-Method-Insight rewriting, 可证伪性/falsifiability checks, reviewer pressure tests, pseudo-innovation diagnosis, or innovation evolution logs for papers, theses, proposals, and technical reports. Paper Novelty Design V1 Use this skill to turn a rough research idea, proposal note, abstract, or introduction draft into a bounded and review-resistant contribution. Prefer this skill when novelty is vague, claims feel inflated, or a paper needs contributions tied to evidence rather than adjectives. Default Workflow Build the minimum research context. Identify field, task, venue or thesis context, current idea, closest known work, evidence, and constraints. If the user provides a draft, extract candidate Problem , Method , and Insight before rewriting. If the user wants a "first" claim or strong literature positioning, state that external literature verification is required. Rewrite the idea as a three-layer novelty statement. Problem : a bounded unresolved gap, contradiction, or limitation. Method : the specific intervention chosen because it addresses that problem. Insight : the portable takeaway the field can reuse if the evidence holds. Reject "our method performs better" as an Insight ; treat it as evidence. Diagnose pseudo-innovation risk. Check for missing motivation, engineering-stack storytelling, domain transplant without adaptation, metric-only narrative, and fragile "first" claims. Read references/review-checks-v1.md when you need the full trap checklist or example rewrites. Convert the core claim into falsifiable propositions. Write 1-3 claims that evidence could support or refute. For each claim, specify support conditions, failure conditions, and the concrete test or evidence needed. If a claim cannot be falsified, narrow its scope until it can. Run pressure tests. Repetition test: has adjacent work already done this? Motivation test: does the method logically target the problem? Falsifiability test: what result would force revision or rejection? Differentiation test: what differs at the Problem, Method, detail, evidence, and Insight layers? Story test: can the contribution be stated in 1 sentence, 3 sentences, and 1 short paragraph? Use references/review-checks-v1.md for detailed prompts and reviewer-style checks. Produce paper-facing output. Give a revised three-layer statement. Give contribution bullets suitable for an introduction, proposal, or thesis section. Give falsifiable hypotheses with a matching evidence plan. Give likely reviewer attack points and concrete repairs. Read references/writing-templates-v1.md when drafting formal contribution statements or evolution logs. Manage iteration honestly. If evidence supports the claim, narrow the claim and state its boundary conditions. If evidence is weak, shrink the scope or recast the contribution as a tested limitation or empirical finding. If experiments reveal a stable unexpected pattern, consider making that pattern the new Insight . Maintain a versioned evolution log with evidence, revision reason, and next falsification test. Output Modes Rough idea or proposal stage ## Novelty Diagnosis Verdict: [strong / promising but underspecified / mostly engineering / too close to prior work] ## Three-Layer Statement | Layer | Draft | |---|---| | Problem | ... | | Method | ... | | Insight | ... | ## Falsifiable Propositions | Proposition | Supporting evidence | Refuting evidence | Required test | |---|---|---|---| | ... | ... | ... | ... | ## Reviewer Pressure Points | Risk | Likely reviewer question | Repair | |---|---|---| | ... | ... | ... | ## Next Revision - ... Mature manuscript or contribution section ## Revised Contributions 1. ... 2. ... 3. ... ## What Each Contribution Must Prove | Contribution | Evidence needed | Weakest link | |---|---|---| | ... | ... | ... | Ongoing project tracking ## Novelty Evolution Log | Version | Current insight | New evidence | Revision reason | Next falsification test | |---|---|---|---|---| | v1 | ... | ... | ... | ... | Resource Routing Read references/review-checks-v1.md for trap diagnosis, falsifiability framing, and pressure-test prompts. Read references/writing-templates-v1.md for three-layer templates, contribution statements, and evolution-log formats. Avoid loading both reference files unless the task spans both diagnosis and final writing. Writing Rules State novelty through bounded claims, not adjectives such as "novel", "advanced", or "state-of-the-art". Treat performance improvements as evidence, not as the novelty itself. Do not invent citations, baselines, datasets, mechanisms, or results. When using "first", narrow the scope and mark it as requiring literature verification. Preserve the user's research direction before proposing a different project. Prefer 3-4 contribution bullets; merge weaker items. Be direct if the idea is mostly engineering or too close to prior work. Keep terminology aligned with the user's source language when rewriting.
Agent 识别该技能的关键词,点击任意一个即可复制。

该技能未提供触发词。

下载的 .skill 包内含以下字段。
字段 说明
format格式标识(skill/v1)
skill_id技能唯一 ID
name技能名称
version版本号
description技能描述
category所属分类(数组)
trigger_words触发词列表
tags标签列表
source来源标识
source_url来源链接(本页地址)
exported_at导出时间(每次下载生成)
system_prompt系统提示词正文
model_config模型参数:provider / model / temperature / max_tokens / top_p
examples示例
install_guide各平台导入说明(Coze / Dify / Claude / 自定义框架)
同一份技能可按不同平台格式导出。
.skill 标准格式,含 system_prompt 与 model_config,导入任意 Agent 框架即可使用 下载
.skillpro 增强格式,额外含脚本 / 工具 / 依赖 / 钩子占位 下载
.json 纯 JSON 导出,只含 system_prompt 与模型参数 下载
Coze 带 frontmatter 的 Markdown,Coze 平台导入用 下载
Dify Dify DSL,创建应用后直接导入 下载

每日精选 Skill 推荐,免费送到你邮箱

输入邮箱,每天接收一个精选 AI Agent 技能推荐。完全免费,持续更新。

提交后我们会发送一封确认邮件,点击邮件里的链接才会开始收信。

完全免费,取消任意时间。我们不会发送垃圾邮件。