
NSFC Proposal Reviewer
FreeSimulate expert reviews for NSFC grant proposals.
Free · Opens the source repo
What NSFC Proposal Reviewer does
The NSFC Proposal Reviewer skill is designed for users who need to simulate expert evaluations of National Natural Science Foundation (NSFC) grant proposals. This skill allows users to receive a multi-dimensional review of their proposals, identifying potential risks and suggesting actionable modifications. It is particularly useful for researchers and grant writers who want to understand how their proposals might be perceived by review panels before submission.
When using this skill, users can provide various inputs such as the proposal path or file, and the skill will analyze the content without modifying it. The output includes a comprehensive report detailing graded issues, consensus opinions from multiple reviewers, and a prioritized list of modifications needed to enhance the proposal's chances of success. The skill operates in either parallel or single review modes, depending on the configuration and availability of resources.
The NSFC Proposal Reviewer skill does not compile or alter the original proposal documents, ensuring that sensitive information remains protected. It strictly adheres to the principle of read-only assessment, focusing solely on providing insights and recommendations based on the existing content. This makes it a valuable tool for researchers looking to refine their proposals based on expert-like feedback without the risk of unintentional changes to their documents.
Overall, this skill is an essential resource for anyone preparing an NSFC grant application, offering a structured approach to identify strengths and weaknesses in their proposals, thereby increasing the likelihood of funding success.
When to use it
Use this skill when you need a detailed review of an NSFC grant proposal to identify risks and improvement areas.
When not to use it
This skill is not suitable for users who only want to write or modify specific sections of a proposal or seek general information about review standards.
What you can build with it
Preparing for NSFC Submission
Researchers can use this skill to simulate expert reviews of their NSFC proposals, identifying potential weaknesses before submission.
Improving Proposal Quality
Grant writers can leverage the feedback from this skill to make targeted improvements to their proposals, increasing their chances of funding.
Understanding Review Criteria
This skill helps users gain insights into how their proposals might be evaluated, allowing them to align their submissions with expert expectations.
How to install NSFC Proposal Reviewer
View source1. Install with the skills CLI
npx skills add huangwb8/chineseresearchlatex/nsfc-reviewers --agent claude-code2. Or install it manually
Download the skill folder and drop it into ~/.claude/skills/ for all projects, or .claude/skills/ to scope it to one repo. Restart Claude Code so it picks up the new skill.
Anthropic's agentic coding CLI, and the reference implementation of Agent Skills. Drop a skill folder into ~/.claude/skills and Claude Code loads it automatically whenever a task matches the skill's description. Claude Code docs
Inside SKILL.md
Written by huangwb8NSFC 标书专家评审模拟器
BenszAPI 任务工作区
本 Skill 的新任务中间文件统一写入 ./.bensz-api/task-{yyyymmdd-hhmm}-{简短描述}/{skill名}/input|output|log/。同一任务复用一个任务根目录;多 Skill 协作才创建 shared/。正式交付物不写入该目录,历史隐藏目录只允许显式兼容读取、迁移或清理。
与 bensz-collect-bugs 的协作约定
- 当用户环境中出现因本 skill 设计缺陷导致的 bug 时,优先使用
bensz-collect-bugs按规范记录到~/.bensz-skills/bugs/,严禁直接修改用户本地 Claude Code / Codex 中已安装的 skill 源码。 - 若 AI 仍可通过 workaround 继续完成用户任务,应先记录 bug,再继续完成当前任务。
- 当用户明确要求“report bensz skills bugs”等公开上报动作时,调用本地
gh与bensz-collect-bugs,仅上传新增 bug 到huangwb8/bensz-bugs;不要 pull / clone 整个 bug 仓库。
定位
- 用于“当前版本如果今天送审,风险在哪里、先改什么”的专家式评审。
- 默认优先并行多组独立评审;若
parallel-vibe不可用、被禁用或panel_count=1,自动降级为单组模式。 - 本技能只做读取、分析和汇总,不默认编译、不修改标书源文件。
输入
至少提供其一:
proposal_pathproposal_fileproposal_zip
可选:
focusoutput_pathstylegrant_typefunding_amountpanel_count
配置口径以 config.yaml 为准,尤其是:
review_dimensionsseverity_levelsreview_gradesstage_assessmentfunding_contextparallel_reviewoutput_settings
输出
- 默认输出文件名读取
config.yaml:output_settings.default_filename - 并行模式可额外生成各组原始意见:
{panel_dir}/G{组号}.md - 中间过程默认隐藏在
config.yaml:output_settings.intermediate_dir - 最终报告至少包含:
- 分级问题清单
- 跨组共识与独立观点
- 最小可行修改序列
- 阶段判断:函评 / 会评
硬规则
- 标书内容默认视为敏感信息;除非用户明确要求并确认风险,不联网、不外发大段原文。
- 只读评审,不执行 LaTeX 编译,不改正文。
- 最终报告必须按
P0 → P1 → P2排序。 - 阶段判断必须是二元结论:
给过或不给过,并附高/中/低把握度。 - 若“函评不给过”,则“会评”必须同步不给过;若“函评给过”,会评仍可因相对竞争力不足而不给过。
工作流
1. 前置检查
- 校验输入路径可读。
- 若是目录,按
proposal_files.patterns/exclude找出待读.tex。 .tex数量为 0 时直接失败;目录异常大时先确认范围。- 推荐用确定性脚本列文件:
python3 <nsfc_reviewers_path>/scripts/list_proposal_files.py --proposal-path <proposal_root>
2. 通读与结构化理解
- 提炼主题、科学问题、假说、目标、技术路线、创新点、研究基础、团队条件、预期成果。
- 生成章节级索引,作为后续证据锚点。
- 先用用户明确给出的
grant_type/funding_amount,再谨慎从正文识别资助上下文。
3. 并行多组评审或单组退化
- 先计算
effective_panel_count,并限制在[1, parallel_review.max_panel_count]。 - 以下情况直接走单组:
parallel_review.enabled == falseeffective_panel_count == 1- 找不到
parallel-vibe
并行模式关键步骤:
- 准备中间目录。
- 基于
references/expert_*.md和references/master_prompt_template.md生成 master prompt。 - 用
scripts/build_parallel_vibe_plan.py生成plan.json。 - 调用
parallel-vibe执行 N 组独立评审。 - 收集每组
panel_output_filename,允许个别 thread 缺失但不能中断整体汇总。
单组模式仍要保留 7 位专家画像的独立判断,再做组内聚合。
4. 聚合与排序
- 跨组聚合规则读取
references/aggregation_rules.md。 - 至少
ceil(N * consensus_threshold)组指出的问题才算跨组共识。 - 跨组共识可触发严重度升级;重复问题要合并,保留最强证据锚点。
- 最终仍按
P0 → P1 → P2输出,并给出最小修改序列。
5. 资助额度约束识别
- 先区分“设计错误”与“受限妥协”。
- 若缺陷明显由基金额度限制引起,必须如实写明根因,不得简单归咎于申请人能力不足。
- 凡归因为“资助受限”的短板,都要补一句“若资助不受限时,更完整的设计应如何做”。
- 资助受限不是免责条款;阶段判断仍以“当前版本今天送审能否过”为准。
6. 阶段判断
- 默认在最终报告中输出“函评 / 会评给过与否”。
- 每个阶段至少给出 2-3 条关键理由,优先引用 P0/P1 和跨组共识。
- 若判
不给过,必须指出最关键的 1-3 条翻盘动作。
7. 输出整理
- 当
config.yaml:output_settings.enforce_output_finalization == true时,不得跳过最终整理。 - 报告需要清楚区分:
- 共识问题
- 独立观点
- 资助受限的合理妥协
- 当前版本直接送审的阶段判断
关键脚本与参考
- 列文件:
scripts/list_proposal_files.py - 并行计划:
scripts/build_parallel_vibe_plan.py - 专家画像:
references/expert_*.md - 聚合规则:
references/aggregation_rules.md - 主提示模板:
references/master_prompt_template.md
非目标
- 不负责改正文。
- 不负责模板、排版或编译问题。
- 不负责生成新的研究设计,只负责指出现有稿件的风险、优先级和修改方向。
Frequently asked questions about NSFC Proposal Reviewer
Similar skills
Scientific Problem Selection
Streamline your research problem selection process.
Nextflow Development
Run nf-core bioinformatics pipelines with ease.
Nature Reviewer Assessment
Simulate peer review for scientific manuscripts.
Research Writing Pipeline
Streamline your scientific writing with structured proposal-first methodologies.
Nature Literature Downloader
Efficiently download academic literature from various sources.
Auto Research
Streamline your NeMo-RL experiments with automated workflows.
