Mathematical Rigor
The `math-rigor` plugin for DeepSeek Harness (DSH) aims to extend agent capabilities through local tools and scripts to address the "hallucination" problem in mathematical proof. The plugin depends on a Python environment, offers 23 MCP tools (covering logic checks, validators, symbolic computation, and number theory), and includes a built-in `/prove` workflow that processes mathematical proofs in a structured, auditable manner. It strictly isolates inputs from a security standpoint and executes only parsed mathematical expressions, ensuring the trustworthiness of the reasoning.
Read Moredsh-math-rigor: A Locally Auditable Mathematical Proving Toolchain
`dsh-math-rigor` is a pluggable architectural component of DeepSeek Harness (DSH), designed to provide automated and auditable mathematical proof capabilities for AI reasoning through a local MCP server and built-in skills. The plugin supports the `/prove` and `/audit-proof` slash commands and executes a six-stage proof workflow, from strategy formulation to machine auditing. It relies on local tools such as Python, SymPy, and Z3 to ensure that reasoning is fully offline and secure, with strictly constrained input execution, making it suitable for agent development scenarios that require high-confidence mathematical derivation.
Read MoreIntroducing the DeepSeek Harness Plugin: dsh-alphasolve
`dsh-alphasolve` is a workflow plugin for DeepSeek Harness (DSH) designed to automate the solving of complex mathematical proofs. Users need to configure a Python/SymPy environment, install the plugin, then place the problem in a `problem.md` file, send a specific prompt, and click the AlphaSolve button to initiate parallel, multi-task solving and verification. Although the tool can assist in developing candidate solutions, users must verify their logical correctness themselves, and the process may be time-consuming.
Read More