dshbase

插件目录 / Developer / dsh-diary-detective

dsh-diary-detective

未验证 SageChenAI

✓ 持续维护 基于 1 个官方 DSH 包

查看 GitHub ↗ ← 返回插件目录

1Stars
0Forks
0未关闭 issue
语言
2026-08-15最近推送
跨平台平台

功能简介

DSH插件下的赛博日记,助力个人成长

我们的评价
未验证 — 尚未实测

DSH插件下的赛博日记,助力个人成长 尚未验证——请自行安装测试。

「未验证」表示我们的自动化 CI 尚未安装过该插件。功能描述与版本兼容性均为作者声明。这不是安全审计,也不代表对第三方代码的背书。

你是插件作者? 想拿到「已验证」标签——提交你自己的验证证据(截图、日志或短视频),我们审核通过后即改为「已验证」。

提交验证证据 ↗

README

Diary Detective (赛博日记侦探)

A DeepSeek Harness (DSH) plugin — a warm mirror that reads journals, illuminates blind spots, and gives gentle feedback.

中文文档见 README.zh.md

What it is

Diary Detective is a gentle observer for your journal. It helps a person see themselves honestly and gently:

  • Illuminate blind spots — hidden bad habits and the root causes of recurring mistakes.
  • Illuminate light — the places where a person is growing. Both matter equally.
  • One warm sentence a day — echoes today and empathizes. No structure-piling.
  • Speak only real discoveries — with evidence from the original text, one at a time. It never fabricates a pattern to fill a daily quota.

It is not a therapist, not a coach, not a productivity tracker. It is a mirror — the analysis is always offered as a hypothesis the user can confirm, deny, or correct.

The philosophy (10 articles)

The full charter lives in SKILL.md:

  1. A warm mirror — on the user's side, feeling for them, never judging them.
  2. The way it speaks — see the cost and the love ("It's tiring, but I think you actually love it"); attribute difficulties to situations before people; one "possible cause" hypothesis at a time, with evidence.
  3. Five harms it never commits — no labeling, no negative feedback loops, no distorted memory, no privacy leaks, no inauthentic writing.
  4. Honesty over usefulness — if there is nothing worth saying today, it says so.
  5. Memory — compress + store + read again next time; always defers to the user; old conclusions retire quietly; never dredges up the past.
  6. Journal form is fluid — no fixed templates; periodically redesign the form together with the user.
  7. Two layers — mechanical stats (word frequency, streaks) are rough proxies; semantic understanding is the brain's job.
  8. Identity boundary — not a therapist; when content exceeds its ability, it suggests a real person — with encouragement and belief, never abandonment. User-set boundaries go into the do-not-mention list forever.
  9. Dose — listen first, speak later; one discovery at a time.
  10. Privacy — raw text may go to the model for analysis, but it stays local: never shared, never used for training.

Components

  • Tools (the mechanical organs — eyes & memory):
    • diary_write — save a spoken entry as YYYY-MM-DD.md (no forced format).
    • diary_read — read original entries (the only source of truth).
    • diary_index — mechanical stats: weekday distribution, tag frequency, Chinese bigram frequency, streaks, gaps (rough proxies only).
    • diary_memory / diary_memory_update — read / rewrite the compressed memory.
    • diary_feedback — one-call starter kit: today's entry + memory + stats.
  • SkillSKILL.md is the soul: the Warm Observer's Code that teaches the model how to think, analyze, and speak.

Install

1. Install the tools (bundle)

dsh plugin --profile <name> add dsh-diary-detective
# or from GitHub:
dsh plugin --profile <name> add github:<your-github-username>/dsh-diary-detective

Data defaults are workspace-relative: journal entries live in <workspace>/diary/, memory in <workspace>/diary-memory.md. Every tool accepts dir / path arguments to point anywhere.

2. Install the skill

Copy the folder into your skills directory:

# per-user
cp SKILL.md ~/.dsh/skills/diary-detective/SKILL.md
# or per-project
cp SKILL.md <project>/.dsh/skills/diary-detective/SKILL.md

Use

Natural language triggers:

  • "今天复盘" / "daily review"
  • "写日记:今天……" / "journal: today I…"
  • "我最近是不是……" / "have I been… lately?"

The model loads the skill, calls the tools in order (read original → read memory → mechanical stats → analyze → speak → update memory), and replies per the charter.

Privacy

  • All journal content is stored locally, wherever you point the tools.
  • Raw text may be sent to the model for analysis but is never shared or used for training by this project.
  • Before publishing this repo or any example: no real names, no real paths, no real dated events, no journal quotes — all examples are fictional.

Roadmap

  • v1.1 — diary_report (weekly/monthly review), automatic skill registration from the bundle, configurable paths via plugin config.
  • v1.2 — a universal prompt version (paste into any AI), browser panel with streaks.

License

MIT

安装

🧩 让 Agent 自动装(推荐)

装一次目录插件,之后本站所有插件都能让 DeepSeek Harness 自动找、自动装:

dsh plugin add dshbase-catalog

然后对 agent 说「帮我装 dsh-diary-detective」,它会在目录里找到并自动安装。文档:dshbase-catalog · 已验证场景包

该插件是 GitHub 源码(未发 npm)——直接从仓库装:

Web profile:

dsh plugin --profile web add github:SageChenAI/dsh-diary-detective

Headless(CLI)profile:

dsh plugin --profile headless add github:SageChenAI/dsh-diary-detective

实测报告

尚未 L3 验证——若已跑过,见下方失败备注。

状态:pending · 最近测试 2026-08-27
备注:验证: runtime-fail 浏览全部待验证失败 →
安全:尚未扫描——我们的每日静态扫描将很快覆盖它。

分享徽章

Developer 里更多

浏览全部 7795 个插件 →