DeepSeek Harness 插件

dsh-habit

Self-learning habit engine for the DeepSeek Harness — detects user-correction signals, judges habits with a low-cost model on threshold, settles candidates behind a two-level human gate(英文原文)

跳到安装方式

来源信息

GitHub 仓库
Max-Null/dsh-habit
最近更新
2026年8月21日
分类
安全与权限
GitHub stars
1
载体类型
plugin
目录证据
上游声明已找到 dsh.bundle
证据路径
package.json#dsh.bundle
核对版本
0.1.0-rc.8
上游核对日期
2026-08-20

该证据由上游目录提供。本站没有安装、运行或安全审核这个插件。

安装

默认先复制一段 Prompt,让 Agent 读 GitHub 仓库和源码;需要自己装时再切到命令。

复制这段 Prompt,发给 DSH、Codex 或其他 Agent,让它先读 GitHub 仓库和源码。

请先不要安装或执行任何命令。阅读这个插件的 GitHub 仓库、README 和关键源码,然后用清楚、直接的方式回答以下问题,帮助我判断它是否适合我的需求:

1. 这个插件是什么,解决什么问题;
2. 适合哪些用户和典型使用场景;
3. 安装后如何使用,并给出一个最小使用示例;
4. 有哪些已知限制,以及隐私、安全、兼容性或维护风险;
5. 给出“推荐 / 有条件推荐 / 不推荐”的明确建议和理由。

请区分仓库明确说明、根据源码推断和未知信息。证据不足时请明确说明,不要猜测或照抄 README。

GitHub:https://github.com/Max-Null/dsh-habit
插件名:dsh-habit
作者:Max-Null

检查来源文件

安装前先看这个插件目录里的 README 和其他文件。

文件资源管理器2 个文件
README.md来源说明 · 只读预览

@max-null/dsh-habit

Self-learning habit engine for the DeepSeek Harness — observes user-correction signals from session events, judges habits with a low-cost model on threshold, and settles candidates behind a two-level human gate. No new agent role: the judgment is an event-driven plugin, immune to context decay.

The loop

① observe   session/event → correction-signal detection (deterministic, zero-token)
② judge     >=3 signals in one session → one flash call (evidence slices + existing habits)
③ settle    candidate zone → user confirms → dsh-memory remember() (suggested)
            → user confirms again → auto → recall injection

Compose

- id: habit
  name: '@max-null/dsh-habit'

Requires storage and llm in the host composition (dsh-base ships both). Installs as a bundle: dsh plugin --profile <name> add @max-null/dsh-habit.

Service

  • ctx.habit — the engine:

- snapshot() → candidates (newest first) - confirm(id) / discard(id) → first-level human gate - (the second gate is dsh-memory's own suggested→auto confirmation)

Config

FieldDefaultMeaning
signalThreshold3Correction signals before one judgment call
providerdeepseek-officialJudgment model provider
modeldeepseek-v4-flashJudgment model (cheap, deterministic)
storageRoot$DSH_HOME/storages/habitJSON storage root

Design notes

  • Deterministic observation, LLM on demand: correction detection is a

fixed phrase list + length cap (task descriptions are not corrections); the LLM only runs when a session accumulates enough signals.

  • Two-level human gate: candidates must be confirmed in the UI AND then

pass dsh-memory's own suggested→auto gate. The model can never promote its own habits.

  • Narrow input for quality: the judgment call gets at most 5 evidence

texts plus the existing habit list — judgment quality comes from precise context, not volume.

Develop

npm install --legacy-peer-deps
npm test
npm run typecheck
npm run build

SSID 系列

本插件属于 **@max-null/* 插件系列——这一系列共同构成 SSID(思灵 · Seek Soul in Darkness)** 桌面体验。SSID 是整合它们的盒:dsh-chat-rail · dsh-chinese-thinking · dsh-draft-polish · dsh-guardian · dsh-habit · dsh-header-unify · dsh-memory · dsh-node-appearance · dsh-plugin-center · dsh-skill-mcp-center · dsh-ssid-panels · dsh-ssid-zh-ui

This plugin belongs to the **@max-null/* family — a set of plugins that together form the SSID (思灵 · Seek Soul in Darkness)** desktop experience.