biociao/dsh-science
112Last commit Aug 15, 2026
dsh-science DSH plugin
dsh-science brings a structured research loop (ReAct state machine) and provenance-tracked artifact management to DeepSeek Harness. It provides 10 science skills covering project setup, literature review, manuscript writing, and bioinformatics. The plugin can be installed as a profile bundle or as an agent preset for a full scientific mode experience.
How to install the dsh-science DSH plugin
dsh plugin --profile web add dsh-science # after npm publishCopying does not run this command. Review the repository and version before installing the dsh-science DSH plugin.
dsh-science DSH plugin data source
dsh-science DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-science DSH plugin can do
- Research loop engine with states: Question → Hypothesis → Experiment → Observe → Analyze → Conclude → Next Question
- Versioned artifacts with SHA-256 provenance, diff, verify, deprecate, and reproduce tools
- 10 science skills including research-loop, science-project-setup, artifact-provenance, scientific-reviewer, literature-connector, parallel-delegation, manuscript-writing, bioinformatics-toolkit, conda-environments, data-inventory
- Concurrency-safe file locking and atomic writes for manifest and artifact files
- Structured error codes (ERR_NOT_INIT, ERR_NOT_FOUND, etc.) and hypothesis state machine (proposed → testing → supported/refuted/inconclusive)
- Manifest-artifact linkage and automatic schema migration (v1→v2)
Where the dsh-science DSH plugin fits
- Conducting reproducible genomics or bioinformatics research projects with a structured loop
- Tracking hypotheses and experiments with versioned artifacts and full provenance
- Collaborating on scientific reviews with automated hypothesis checking and report generation
- Setting up new projects with a standardized skeleton (experiments, literature, artifacts, etc.)
- Integrating DSH with conda environments and bioinformatics toolkits via skills
Who the dsh-science DSH plugin is for
- Bioinformaticians and genomics researchers using DeepSeek Harness
- Scientists who want a systematic, reproducible research workflow inside an AI agent
dsh-science DSH plugin limitations
- Requires DSH (DeepSeek Harness) version 0.1.0-rc.6 or later and Node.js >=18
- Early stage (v0.1.1 hardening); some features may still be evolving
dsh-science DSH plugin: from the repository README
Quoted from the biociao/dsh-science README, the upstream source of the dsh-science DSH plugin. Copyright remains with the original authors.
**A Claude Science–style research workbench for [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) — for genomics / pathogens / human health / bioinformatics projects.** > One-liner: **dsh-science** — Claude Science-style research workbench for DSH: ReAct research-loop engine (research_* tools), versioned artifacts with provenance (artifact_* tools), and 10 science skills for genomics / pathogens / bioinformatics. - **ReAct research loop engine** — `research_init` / `research_state` / `research_hypothesis` / `research_experiment` / `research_findings` / `research_phase` / `research_review` / `research_report`, persisted in a `research-manifest.json` state machine (Question → Hypothesis → Experiment → Observe → Analyze → Conclude → Next Question). - **Versioned artifacts with provenance** — `artifact_save` / `artifact_list` / `artifact_show` / `artifact_diff` / `artifact_verify` / `artifact_deprecate` / `artifact_reproduce`: every result saved as `artifacts/<name>/v<N>/` with per-file SHA-256, `artifact.json` provenance (command / inputs / environment / envFile) and an append-only `provenance.md`. - **10 science skills** — research-loop, science-project-setup, art
Read the full READMERepository license: MIT
dsh-science DSH plugin questions
How do I install dsh-science as a DSH plugin?
You can install it as a profile bundle with `dsh plugin --profile web add dsh-science` (after npm publish) or directly from GitHub with `dsh plugin --profile web add "github:biociao/dsh-science"`. After that, restart the profile or refresh the Web GUI to make the tools available.
What is the difference between the bundle and the agent preset installation?
The bundle installation makes the `research_*` and `artifact_*` tools available to every agent on the profile with a single command. The agent preset (科学模式) provides the full scientific mode experience with a research persona and per-agent scoping; it requires cloning the repo and running `scripts/install.sh`.
How do I start a research session with dsh-science?
First run `research_init` to create the project skeleton and `research-manifest.json`. Then at the start of each session, read `research_state` to see the current loop state. Use `research_hypothesis`, `research_experiment`, `research_findings`, and `artifact_save` to run the loop and save results.
Where do the 10 science skills come from and how do I install them?
The skills are provided in the `skills/` directory of the repo. They are auto-discovered from a project's `.dsh/skills/` folder. You can install them machine-wide by running `bash scripts/install-skills.sh`, which copies them to `~/.dsh/skills` (respecting `$DSH_HOME`).
Does dsh-science support concurrency and prevent data corruption?
Yes, version 0.1.1 includes concurrency-safe file locking (O_EXCL + stale reclaim) and atomic writes (tmp+rename) for both the research manifest and artifact files. Parallel subagents can no longer corrupt or lose updates on `research-manifest.json` or `artifacts.json`. The plugin also uses structured error codes to help debug issues.