beibingyangliuying/dsh-autoresearch

Autoresearch is a domain-neutral research workbench for DeepSeek Harness: a selectable agent preset plus a set of on-demand research disciplines and a scaffold for one study. It constrains how research is conducted - evidence typing, challenge, handoff, run records and settlement - while saying nothing about what is researched, and the repository root is itself the bundle, needing no build step.

Autoresearch is a domain-neutral research workbench for DeepSeek Harness: a selectable agent preset, a set of on-demand research disciplines and a scaffold for one study. It constrains how research is conducted - evidence typing, challenge, handoff, run records and settlement - and deliberately says nothing about what is researched. Its contract is normative in protocol.md, which wins over the README, the composition file or any discipline text, and the symbol names live in a single index rather than being repeated. The GitHub install path is verified end to end: the package lands in the profile as a real copy, pnpm hoists the one dependency so the row's import resolves, and the composition then carries the preset and its three rows.

Learning & Research ★ 0 updated 2026-10-13
View on GitHub ↗

Install

dsh plugin --profile web add github:beibingyangliuying/dsh-autoresearch

Copied verbatim from the README's Install block, which notes the repository root is itself the bundle and needs no build step; confirm registration afterwards with dsh --profile web --dump-config | Select-String 'preset-autoresearch'.

Compatibility

A DSH bundle: a package whose package.json points at a composition patch, so installing the package registers the preset. The preset must be selected where the new session is created - switching it inside an existing session does not re-create it. English README.

Details

Alternatives

1692775560/dsh-Mimir-Academic-research · Astervolans/dsh-literature-search · benz-ai-x/dsh-research-graph

More plugins in Learning & Research

Browse more in Learning & Research

Guides for Learning & Research plugins