DeepSeek Harness Plugin Hub

Publish and manage complete Harness Profiles. Discover Plugins for your next setup.

Explore

PluginsPresetsDocsNews

Community

Publish a pluginContactReport an issue

Resources

Plugin Hub on GitHubDeepSeek HarnessSystem statusPrivacy notice
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

Independent and unofficial. Not affiliated with, authorized by, or endorsed by DeepSeek.

Gatecraft — DSH Plugin for DeepSeek Harness
DeepSeek Harness Plugin Hub
ProfilesPluginsCategoriesNewsDocsSign inManage Profiles
ProfilesPluginsCategoriesNewsDocsSign in
← Plugins
G

gatecraft

Gatecraft

GateCraft: A gated mathematical modeling skill suite for DeepSeek Harness—five-stage pipeline, constraint classification (facts/invariants/observations), national-award writing conventions, and mechanical acceptance via docgate; agent-based solution quality checks, with humans thinking and making de

The plugin will be installed here. Keep web if you are unsure.

npx -y @deepseek-ai/dsh plugin --profile web add github:Crayonnan/dsh-math-modeling-skills-Gatecraft-#b4dfb3eeb3c8e421056132da940f5d78b682aa25
READMECompatibilityVersions

Description

GateCraft: A gated mathematical modeling skill suite for DeepSeek Harness—five-stage pipeline, constraint classification (facts/invariants/observations), national-award writing conventions, and mechanical acceptance via docgate; agent-based solution quality checks, with humans thinking and making decisions at stage gates.

Compatibility and provenance

Gatecraft is published as gatecraft and currently resolves to version 0.3.0. The Hub verifies its manifest and preserves the exact installation source for reproducible installs.

DSH compatibility
*
Runtime surfaces
any
Release source
github
Registry updated
9/8/2026

Versions

0.3.0stable
9/8/2026
0.2.0stable
8/24/2026
0.1.0stable
8/20/2026

Related plugins

Loading related plugins…

Latest
0.3.0
DSH
*
HMR
Process restart
Tree shaking
Safe tree shaking not declared
Unpacked size
Unavailable
Files
Unavailable
Surface
any
License
MIT
Source
github
GitHub
★ 0
Weekly downloads
0
View source ↗
README badge

Click the badge to copy Markdown for your README.

Do you maintain this Plugin?Claim benefit · Priority security scan

Verify the GitHub repository declared in package.json to manage this listing. After you claim it, Hub will prioritize a security scan of the current version and publish the result when it passes.

Claim this Plugin →
Report an issue

Related plugins

More verified plugins in developer-tools.

Client Ui Git Graph@linxin666/dsh-client-ui-git-graphExternal dsh web GUI plugin: a blank-session git branch selector + Git graph, with real host-side git operations and guards, as a dsh profile bundleDoctor@linxin666/dsh-doctorTransactional rescue mode for DSH profiles with a supervised launcher, isolated recovery capsule, deterministic repairs, health monitoring, and a local Web recovery consoleSsh@linxin666/dsh-sshRemote SSH operations for the dsh web GUI: host config store (~/.dsh/dsh-ssh.json, import from ~/.ssh/config), a persistent ssh2 connection pool with jump-host support, exec / PTY web terminal / SFTP transfer / local port-forward tunnels / cluster executiFind Plugindsh-find-pluginFind DeepSeek Harness plugins inside the agent — live GitHub dsh-plugin topic search, ranked by stars.

README

GateCraft(门控工艺)

English

门控式数学建模 skill 套件(9 skills + DSH preset)for DeepSeek Harness。不做无脑端到端自动化——agent 求解与质检,人在每个阶段门思考与决策,产出带自己品味的建模成果。

内容

  • 9 skills:competition-workflow(五阶段流水线总控:阶段报告 / EDA 五问 / 验证三件套)· guozhan-paper(国奖写作范式)· vision-ocr(题面与范文阅读)· sensitivity-analysis · statistical-diagnosis · math-modeling-paper(论文内容)· math-paper-template(LaTeX 排版)· tex-pdf-image-to-word(转换 Word)· paper-gate(交付验收层 + 全仓库约束分级约定的唯一真源)
  • 按需加载:competition-workflow/references/(模型选型决策树、交稿自查)与 guozhan-paper/references/(获奖论文实证语料、页码级正反例)不占常驻上下文,用到时才读。
  • assets:optimization-playbook(优化求解/验证决策表)· figure-playbook(流程图与图件模板)· prompt-pack(14 条实战提示词)· flowchart_gen.py(规格 → drawio 生成器)· ocr_batch.py(并发 OCR)· docgate.py(paper-gate 执行引擎)· official-paper-format.md(官方格式真源)
  • DSH preset:presets/math-modeling/ — 粘贴一道竞赛题即可自动启动工作流

安装

dsh plugin add Crayonnan/dsh-math-modeling-skills-Gatecraft-

可选:把 presets/math-modeling/ 拷贝到 ${DSH_HOME:-$HOME}/.dsh/.agent-presets/math-modeling/,新会话选择"数学建模模式"。

技能协作图(数模一条龙)

                    competition-workflow(总控,两模式:流水线 / Day1-3 时间线)
                              │ 阶段0 读题
                              ▼
                         vision-ocr(OCR 落盘,按需取段)
                              │ 阶段1 分析建模 → ANALYSIS_MODELING_REPORT.md
                              │ 阶段2 代码结果 → RESULTS_REPORT.md
                              │     └─ statistical-diagnosis(模型诊断 → DIAGNOSIS_REPORT.md)
                              │ 阶段3 灵敏度 → sensitivity-analysis(题型自适应 → SENSITIVITY_REPORT.md)
                              │ 阶段4 论文
                              ▼
          math-modeling-paper(内容) ──► math-paper-template(排版) ──► PDF
                 │ 参考 guozhan-paper(国奖范式)       │ 需要 Word 版
                 └── official-paper-format.md ◄────────┴──► tex-pdf-image-to-word(11 条检查)
                              │ 阶段5 验收 → VERIFY_DOCGATE.md + VERIFY_REPORT.md
                              ▼
                  paper-gate + docgate.py(机械检查:FAIL 只拦事实错误与真实性问题)
                              ▼
                  references/submission-checklist.md(docgate 查不到的人工自查问句)

paper-gate —— 交付验收层与约束分级真源

模型内核达标而失分集中在表达层与验收层(图表超版心、编号双轨、AI 工具混入文献、摘要数量级跳变、声明强度超过证据等级、篇幅失衡、错别字群),且既有检查全部绑定 LaTeX 管线、对实际提交的 docx 工件静默失效。由此建立:

  1. 约束分级(全仓库唯一说明处):规则分三层——【事实】官方明文与工程事实、【不变量】可机械验证的真实性与一致性、【观察】获奖样本统计。前两层可以硬,且硬在代码里;第三层写硬了会诱发凑指标,只能写成"样本中常见 + 以当届模板为准"。
  2. 工件唯一性:只认最终提交文件;任何转换/另存后重跑检查(源文件通过 ≠ 提交版通过)。
  3. docgate.py 机械检查:对 docx(OOXML)/tex 双后端执行 13 项——图片几何(版心动态读取)、图/表/式编号对账、AI 痕迹扫描、摘要数量级哨兵、篇幅均衡、跨章重复、错别字模式库、变量空解释残骸、重述原创度等。FAIL = 事实错误或真实性问题(编号断裂、摘要数字在正文找不到、图片超版心…),修复后重跑;WARN = 手艺判断,逐条人工裁决。 报告头披露 WARN/SKIP 数量,防止把"gate 通过"读成"论文没问题"。
  4. 声明强度校准:求解结论按证据等级五档(解析证明→仅启发式)映射允许措辞。
  5. 规则参数化:阈值/词表/赛事页数与 FAIL/WARN 归属全部在 paper-gate-rules.yaml,换赛事只改配置、不改 docgate.py。

用法:python assets/docgate.py <提交文件.docx|.tex> [--results RESULTS_REPORT.md] [--problem 题面.txt]

依赖与回归测试:pip install -r requirements.txt 后运行 python tests/test_docgate.py。回归输入是 tests/make_fixtures.py 合成的带已知缺陷夹具(私有论文不入库),除判级外还锁定一组基线:脏文档 5 FAIL(02/03 文献区/04 数量级/08/12)+ 5 WARN(01/03 正文/05/06/09),干净文档 0 FAIL。

理念

  • 硬约束只留给事实与真实性:官方明文、工程约束、可机械验证的一致性用硬语气,且尽量写进 docgate.py 而不是散文;篇幅、句式、检验选型这类手艺判断用"常见误区 + 推荐/不推荐 + 自查问句"表达。对强模型来说,把手艺分歧伪装成阻断条件不会更严格,只会诱发凑指标。
  • 阶段报告不是审批关卡:它的作用是让下一阶段的每个数字有出处。一个子问题通常迭代 2-3 轮,每轮记录"改动 → 效果 → 指标";指标不达标时,是继续优化还是如实声明局限,由人判断。
  • 报告先行:论文每个句子都来自阶段报告中的事实;范文句子只作样例,不抄模板。
  • 数值纪律:论文每个数字只许来自 reports/ 报告或代码输出;改脚本重跑后做「论文数字↔csv」零漂移核对。
  • 批判性验证:外部指南逐条核实、第三方结论重算、结果对照文献基准。
  • 品味来自范式:四项衔接要求(R1-R4)每条带"标准 + 正面样例(含页码) + 反例(含页码)"。

适用范围

在统计分析类与优化/决策类(典型" C"题)上经过实战检验。机理/物理仿真(A 题)与图论/工程(B 题)未经检验——自行扩充检查清单并回馈社区。

与 MathModelAgent 的分工

不是重复,是分工:其求解器作为后端(mma_exec_python 钩子已预留),GateCraft 是编排与质检层——思考、转向与深度参与发生在阶段门上。

结构

skills/         9 skills(competition-workflow 为总控)
assets/         playbooks / prompt-pack / docgate.py / 生成器(与 skills 同步)
presets/        math-modeling(DSH preset)
index.js + cordis.patch.yml + package.json   dsh bundle 打包

License

MIT。贡献遵循一种格式:要求 / 可判定标准 / 正面样例(含页码) / 反例(含页码)——每条清单项必须来自一次真实失败或一次真实获奖。