BSD-3-Clause DSH writing plugin with eight deterministic tools — AI-pattern scan, voice fingerprint learning, 0–100 scoring — that return rewrite briefs instead of calling a model.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-11-21
License
BSD-3-Clause
01What can it help you accomplish?
Strip generic AI-sounding patterns from a draft inside a DeepSeek Harness session
An AI-ness score with the matched rules from `humanize_scan`, then a `humanize_rewrite` brief (rules + issues found in this text) that the agent applies itself
Writers and developers running DSH who want agent-edited copy without the usual AI boilerplate
Teach DSH your own writing fingerprint and score drafts against it
A voice profile persisted to `~/.dsh/voice-profiles.json` (fingerprint + samples) and a 0–100 similarity score with a per-feature breakdown
People who publish under their own name and want AI assistance that still sounds like them
Rewrite in a learned voice without spending another model call on analysis
A `voice_rewrite` brief (fingerprint + few-shot samples + issues found); the agent does the rewriting in its own turn, then `voice_score` shows the before/after gap
DSH users who want reproducible, inspectable editing instead of a black-box LLM rewrite pass
02How to install into DeepSeek Harness
Prerequisites
- DeepSeek Harness with the `dsh` CLI available — the plugin is added with `dsh plugin add`
- No extra runtime dependency: the GitHub install ships a prebuilt `lib/`, and the bundle is self-contained with no unpublished shared dependency
Installation steps
- 01
Run `dsh plugin add github:lynote-ai/dsh-humanizer` — the README's GitHub install path, which ships the prebuilt `lib/`
$ dsh plugin add github:lynote-ai/dsh-humanizer
- 02
Optional: add an insert block for `dsh-humanizer` with `strength: standard` (`light` | `standard` | `aggressive`), `storagePath: ~/.dsh/voice-profiles.json` and `maxExcerpts: 3` — every field has a default
Verify the integration
Not specified by the author
03DSH integration and capability boundaries
Native DSH plugin added with the `dsh plugin add` CLI command; it registers eight deterministic `humanize_*` / `voice_*` tools that the agent calls inside a DSH session, with an optional `config` block (strength / storagePath / maxExcerpts).
De-AI scan
any draft text→an AI-ness score plus the specific rules that matched
Rule catalogue export
the full AI-writing pattern catalogue, readable and editable
De-AI rewrite brief
a draft plus the scan findings→a brief of rules and issues; the agent performs the rewrite in its own turn
returns a brief only — the actual rewriting happens in the agent turnVoice profile management
writing samples, a profile name→a stored fingerprint profile; `voice_profile` reads one or lists all, `voice_remove` deletes one
writes profile JSON to the configured storage path (default `~/.dsh/voice-profiles.json`)`voice_remove` deletes a stored profile file entryVoice similarity scoring
text plus a profile name→a 0–100 similarity score with a per-feature breakdown (sentence length, punctuation, stance, vocabulary, lexical richness)
Voice rewrite brief
a draft plus a profile name→a brief with the fingerprint, few-shot samples and issues found
returns a brief only — no model call, the agent rewrites
04Who is it for? When not to use it?
Good for
- Writers and developers running DSH who want agent-edited copy without the usual AI boilerplate
- People who publish under their own name and want AI assistance that still sounds like them
- DSH users who want reproducible, inspectable editing instead of a black-box LLM rewrite pass
05Compatibility, maintenance and safety notes
- The plugin never calls a model: `humanize_rewrite` and `voice_rewrite` only return a brief, so the rewriting still depends on the agent applying it in its own turn.
- The npm install path (`dsh plugin add dsh-humanizer`) is only documented as "once published to npm" — today the GitHub form is the working install.
- Voice profiles are plain files: samples and fingerprints are written under `~/.dsh/` (default `voice-profiles.json`, configurable via `storagePath`) with no storage backend.
- The AI-writing rule catalogue covers English and Chinese only, modeled on stop-slop / Humanizer-zh.
BSD-3-Clause · no releases yet (repo created 2026-08-20, last push 2026-08-20)
06Frequently asked questions
How do I install dsh-humanizer in DeepSeek Harness?
Run `dsh plugin add github:lynote-ai/dsh-humanizer`. That GitHub path ships a prebuilt `lib/`, so there is no build step on your side. The shorter `dsh plugin add dsh-humanizer` is only noted as available once the package is published to npm.
Where do the tools show up after install, and how do I trigger them?
They are agent tools, not slash commands or an LLM wrapper: inside a DSH session ask for the job in plain language — "scan this draft for AI patterns", "build a voice profile from these samples", "score this against my profile" — and the agent calls `humanize_scan`, `voice_import`, `voice_score` or `voice_rewrite` on its own.
Does it burn model calls or tokens?
No. All eight tools are deterministic and perform no model calls; they return rules, fingerprints, scores and briefs, and the actual rewriting happens in the agent's normal turn. That also makes the output reproducible and unit-testable.
Does it need a specific DSH version or platform?
The README states no version floor, OS restriction or extra dependency: the bundle is self-contained and profiles are plain JSON under `~/.dsh/`. A working `dsh` CLI with plugin support is all the install section assumes.
How does it sit next to other DSH plugins?
It wraps no model provider — it only produces rules, fingerprints, scores and briefs — so it can run alongside whatever model or other plugins your DSH session already uses. The one shared surface is `~/.dsh/`, where its voice profiles are stored.
07Related DSH workflows
dsh-market
by dsh-market
插件市场 (plugin marketplace): 打开 Settings → Plugin Market 浏览/搜索/一键安装社区插件;安装源限制为 curated awesome-dsh-plugin 注册表(其余拒绝),默认禁用构建脚本(pnpm>=10 需显式开启)。
dsh-guizang-design
by op7418
Expose five content/design skills, a gallery, and multi-skill workflows inside DSH.
oh-story-dsh
by zenstory-ai
Fiction/short-drama writing workbench: 23 skills, roles, markdown/JSONL preview, live file-follow.
dsh-skills-manager
by michengai
DSH 技能集合管理器,统一管理多个 skill。
08Data and sources
A writing skill for [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) (`dsh`).
dsh plugin add github:lynote-ai/dsh-humanizer
- **Single self-contained bundle** — no unpublished shared dependency, so `dsh plugin add` just works.
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-11-21. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.
