dsh-plugin-dreaming 1.0.1

dsh-plugin-dreaming 1.0.1 — DSH 0.2.1-rc.3 compatibility re-verification

A compatibility pass against DeepSeek Harness Desktop 0.2.1-rc.3, plus four pre-existing defects the
re-check surfaced.

Compatibility

• Re-verified live on 0.2.1-rc.3: the seven Host packages the plugin depends on (dsh-tools,
  dsh-system-prompt, dsh-skill, dsh-session, dsh-agent, dsh-llm, dsh-settings) are
  byte-identical to 0.2.1-rc.2; all 21 Host extension points, the dashboard API, the client module
  loader ABI and the settings.section slot still hold. No adaptation change was required.
• cordis / loader unchanged; client bundle changes between rc.2 and rc.3 are build artifacts, not
  semantic drift.

Fixes (pre-existing, not rc.3 regressions)

• Prompt section interpolation — the capability-awareness section now registers with
  interpolate: false. Its text is generated prose (skill names, judge/dream free text) and could
  otherwise trip the Host's strict {{variable}} interpolation and fail prompt assembly.
• api.enabled — setting it to false now actually skips mounting /dreaming/api/* instead of
  only changing the reported config.
• Panel language — the client reads the active locale id from snapshot.active (falling back to
  the retired snapshot.language), so the panel follows the in-app language setting rather than the
  browser language.
• Docs — the architecture baseline moved to 0.2.1-rc.3, the documented peer range now matches
  package.json (>=0.2.1-0), and the raw-parameters / output.schema comment matches ToolRuntime.

263 unit tests (two new regression cases).

---

dsh-plugin-dreaming 1.0.1 — DSH 0.2.1-rc.3 兼容性复核

针对 DeepSeek Harness Desktop 0.2.1-rc.3 做了一轮兼容性复核，并修掉复核中暴露的四个既有缺陷。

兼容性

• 在 0.2.1-rc.3 上实测复核：插件依赖的 7 个宿主包（dsh-tools、dsh-system-prompt、
  dsh-skill、dsh-session、dsh-agent、dsh-llm、dsh-settings）与 0.2.1-rc.2 字节相同；
  21 个宿主扩展点、面板 HTTP API、客户端模块加载 ABI、settings.section 槽位全部成立。
  不需要为适配新版本改代码。
• cordis / loader 未变；rc.2→rc.3 客户端 bundle 的差异是构建产物，不是语义变化。

修复（均为既有问题，不是 rc.3 回归）

• prompt 段落插值：能力自省段落现在注册 interpolate: false。该段文本是生成内容（技能名、
  judge/dream 自由文本），否则可能触发宿主严格的 {{variable}} 插值，导致 prompt 组装失败。
• api.enabled：设为 false 时现在真的不挂 /dreaming/api/*，而不是只改展示的配置。
• 面板语言：客户端改读 snapshot.active（旧字段 snapshot.language 兜底），面板跟随应用内
  语言设置，而不是浏览器语言。
• 文档：架构基线更新到 0.2.1-rc.3，文档里的 peer 范围与 package.json 对齐（>=0.2.1-0），
  raw parameters / output.schema 的注释与 ToolRuntime 实际行为一致。

263 个单元测试（含 2 个新回归用例）。

dsh-plugin-dreaming 1.0.0

Dreaming: measurable, correctable "consciousness strength" for skills

dsh-plugin-dreaming is a global DeepSeek Harness plugin (Host + Web panel) for a concrete problem:
a skill description is a claim, and claims routinely disagree with delivery. The model loads the whole
body, then discovers the skill cannot do the job — tokens wasted.

It is not a logging plugin; it is a capability self-model:

• every skill load is concurrently probed with a cheap call that predicts what the skill can really
  deliver for this task, where it will mislead, and whether a cheaper path exists;
• it observes what actually happened (followed the skill, abandoned it, redid the work, token cost);
• at turn end a judge compares expectation with reality and extracts reusable rules;
• each skill keeps a capability belief plus a consciousness score (fit accuracy / boundary accuracy /
  precision / waste rate);
• the measured truth is injected before use (system-prompt section + skill-result annotation);
• a nightly dream revisits mismatches and safely edits user-level skill descriptions and boundary
  sections — backup first, journal, hash verification, conflict refusal, single-step revert.

No edits by default: dream.autoApply = false records suggestions only; enable it explicitly to let
dreams apply changes. The panel lives in Settings → Dreaming and shows consciousness, per-skill measured
history, recent calls, the dream journal and revert controls.

Install (DSH ≥ 0.2.1, from GitHub):

dsh plugin --profile <profile> add git+https://github.com/puppetcat-fire/dsh-plugin-dreaming.git

• Source and docs: https://github.com/puppetcat-fire/dsh-plugin-dreaming
• MIT; 261 unit tests; no test-framework dependency.
• Not published to npm yet; install straight from Git (npm publishing is planned).

Try it and tell us if the measured consciousness disagrees with your intuition — that is exactly the
signal we want.

---

Dreaming：让 skill 的“意识强度”可测量、可修正

dsh-plugin-dreaming 是一个 DSH 全局插件（Host + Web 面板），解决一个具体问题：skill 的描述是
“声明能力”，但声明和实际交付经常不一致——模型加载了整段正文，才发现它做不了这件事，token 白花。

它不是日志插件，而是能力自模型：

• 每次 skill 加载时并发做一次便宜的 probe，预测它对这个任务真正能交付什么、哪里会踩坑、
  有没有更便宜的替代路径；
• 观察加载之后真实发生的事（是否按 skill 执行、是否放弃/返工、token 代价）；
• 回合结束后 judge 预期 vs 现实，产出可复用规则；
• 每个 skill 维护能力信念与意识强度指标（fit 准确率 / 边界准确率 / 精确率 / 浪费率）；
• 在模型使用 skill 前，把实测真相注入上下文（系统提示段落 + skill 结果注解）；
• 夜间做梦：回到不匹配案例，安全地修正用户级 skill 的描述与能力边界——先备份、写日志、
  哈希校验、冲突拒绝、单次回滚。

默认不改文件：dream.autoApply = false，做梦只记录建议；要自动修正需显式打开。面板在
Settings → Dreaming，可查看意识强度、每个 skill 的实测历史、最近调用与梦境日志，并手动回滚。

安装（DSH ≥ 0.2.1，从 GitHub 安装）：

dsh plugin --profile <profile> add git+https://github.com/puppetcat-fire/dsh-plugin-dreaming.git

• 源码与文档：https://github.com/puppetcat-fire/dsh-plugin-dreaming
• MIT；261 个单元测试；无测试框架依赖。
• 目前未发布到 npm，直接从 Git 安装（npm 发布计划中）。

欢迎试用并反馈：如果它测出的“意识强度”和你的直觉不一致，那正是我们想知道的。