A native DSH mathematical-modeling plugin with five role presets, a bundled model library, staged verification, and paper-review gates.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-09-13
License
MIT
01What can it help you accomplish?
Run a complete mathematical-modeling competition workflow in DSH
Problem framing, derivation, coding, independent verification, paper writing, and review across one preset
DeepSeek Harness users solving mathematical-modeling competition problems
Work on a single modeling stage or resume an existing solution
Stage-specific modeling, coding, writing, or review work plus checkpoint-based continuation
Users who need focused intervention or want to continue a DSH workspace from a saved progress file
Apply quality gates and multi-perspective review before finalizing a paper
Phase gates, independent verification, blocking-item checks, and three-seat structured review
Teams that want explicit checks around reproducibility, correctness, and academic integrity
02How to install into DeepSeek Harness
Prerequisites
- A DeepSeek Harness installation with the `dsh` plugin CLI
Installation steps
- 01
Run `dsh plugin --profile web add github:Andiii208/dsh-ultramath`.
$ dsh plugin --profile web add github:Andiii208/dsh-ultramath
- 02
Restart DSH, then create a new session and choose the UltraMath preset.
Verify the integration
- After restarting DSH, confirm the preset selector shows `UltraMath` plus the four single-stage roles.
- Confirm the skill selector shows `ultramath` and the model library path is `~/.dsh/skills/ultramath/模型库/`.
03DSH integration and capability boundaries
Native DeepSeek Harness plugin with an agent preset and skill bundle installed through the dsh plugin CLI.
Five-role agent preset bundle
A mathematical-modeling problem and its data attachments→One UltraMath controller plus mathematician, engineer, writer, and reviewer entry points
The plugin synchronizes presets into the DSH agent-preset directory at startupModel library and paper toolchain
A modeling problem and the relevant method or paper task→A bundled model-library index, paper templates, and review scripts
The package synchronizes its bundled skills, templates, and scripts into `~/.dsh/skills/ultramath/`Independent verification and review gates
Phase outputs and a draft mathematical-modeling paper→Independent checking, blocking-item review, and multi-seat scorecard review
Review findings can send the workflow back to an earlier phaseProgress and reproducibility controls
Code outputs, CSV files, and a paper draft→Progress artifacts, frozen numbers, stale-reference checks, and reproducibility-oriented gates
The workflow expects code execution to produce CSV and PNG result artifacts
04Who is it for? When not to use it?
Good for
- DeepSeek Harness users solving mathematical-modeling competition problems
- Users who need focused intervention or want to continue a DSH workspace from a saved progress file
- Teams that want explicit checks around reproducibility, correctness, and academic integrity
Not for
- The README advertises Node.js >=22 for the package.
- The documented paper phase uses xelatex twice and requires zero errors.
05Compatibility, maintenance and safety notes
- The README advertises Node.js >=22 for the package.
- The documented paper phase uses xelatex twice and requires zero errors.
- The README's full-automatic quick start is described as running without questions and completing in one pass; users can instead select Friendly Mode for confirmations at key decisions.
MIT · latest release v0.6.0
06Frequently asked questions
What does dsh-ultramath do in DeepSeek Harness?
It packages a mathematical-modeling workflow as a native DSH agent preset and skill bundle, covering problem reading, framing, derivation, coding, verification, paper writing, and review.
How do I install it?
Run `dsh plugin --profile web add github:Andiii208/dsh-ultramath`, restart DSH, create a new session, and choose the UltraMath preset.
Which DSH roles are included?
The README describes the UltraMath controller plus four single-stage roles: mathematician, engineer, writer, and reviewer.
Can I run only one stage or continue later?
Yes. The README gives examples for doing only modeling and solving, writing and compiling a paper, or running review convergence. It also describes continuing from `求解/进度.md`.
What should I check before treating a paper as finished?
The documented flow includes independent verification, Phase 3 xelatex compilation with zero errors, B1-B9 blocking checks, and three-seat blind review.
07Related DSH workflows
archify
by tt-a1i
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.
openviking
by volcengine
Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
nocobase
by nocobase
NocoBase is an open-source AI + no-code platform for building business systems fast. Instead of generating everything from scratch, AI works on top of production-proven infrastructure and a WYSIWYG no-code interface, so you get both speed and reliability.
learn-harness-engineering
by walkinglabs
Harness engineering beginner tutorial, from 0 to 1
08Data and sources
把获奖级数模方法论打包成 DeepSeek Harness 原生 agent 预设 + 技能包。
dsh plugin --profile web add github:Andiii208/dsh-ultramath
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-09-13. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.