Soul Keeper

监测 workspace files 的更新时机，在正确的时候提示优化。基于 WORKSPACE_FILES_GUIDE.md 执行。 **自动触发**： - 用户纠正同类行为 2+ 次 → SOUL.md 该更新了 - 用户表达新偏好/新习惯 → USER.md 该更新了 - 新项目启动或方向变化 → MEMO...

Security Scan

The skill's instructions, scope, and lack of required credentials/install steps align with its stated purpose of monitoring and suggesting edits to workspace documentation files.

Purpose & Capability

Name/description match the behavior in SKILL.md: the skill monitors workspace files (SOUL.md, USER.md, AGENTS.md, etc.) and proposes updates. It does not request unrelated binaries, credentials, or config paths, so the requested capabilities are proportionate to the stated purpose.

Instruction Scope

SKILL.md explicitly instructs the agent to read docs/WORKSPACE_FILES_GUIDE.md and the named workspace files, analyze differences, and propose edits; it will only write after explicit user confirmation. This is consistent with the purpose but does imply the skill will read project files (which could contain sensitive data if present). The SKILL.md does not instruct access to system-wide paths, secrets, or external endpoints.

Install Mechanism

Instruction-only skill with no install spec and no code files — lowest-risk install posture. Nothing is downloaded or written during installation.

Credentials

No environment variables, credentials, or config paths are required. The amount and type of access requested (read/write to workspace files after confirmation) are appropriate for a workspace-file management helper.

Persistence & Privilege

always:false and no special privileges requested. The skill can be invoked autonomously by the agent (platform default), which fits its proactive-checking design; it does not request permanent platform-wide presence or modify other skills' configs.

Assessment

This skill appears coherent and does what it claims: read workspace documentation files, propose a single edit at a time, and only apply changes after you confirm. Before installing: (1) review docs/WORKSPACE_FILES_GUIDE.md so you know its edit rules; (2) ensure you don't store secrets or private credentials in the workspace files the skill will read; (3) be aware the agent may proactively suggest updates (autonomous invocation is allowed by default), but it will not auto-apply changes without your confirmation.

Workspace Optimizer

核心理念

每个文件有自己的自然更新节奏，不是等用户抱怨才改。

各文件的 Trigger 表

文件	更新时机	信号
SOUL.md	Agent 行为需要调整	同类纠正 2+ 次；沟通风格变了；新边界
USER.md	用户情况变了	新方向；新偏好；新习惯；新设备/环境
AGENTS.md	工作流程变了	新 docs/ 规则文件；启动流程调整；新触发规则
TOOLS.md	工具链变了	新工具；旧工具出事；致命用法发现
MEMORY.md	世界状态变了	新项目；项目完结；新环境配置
WORKING.md	项目进度变了	实质进展；新项目；项目暂停/完成
IDENTITY.md	几乎不变	改名；改定位
HEARTBEAT.md	检查清单变了	新定时任务；旧任务移除

触发后流程

1. 读写作指南

必须先读：docs/WORKSPACE_FILES_GUIDE.md

2. 读取对应文件

根据 trigger 类型读对应文件。

3. 对比分析

• 文件内容是否和现实一致？
• 该加的信息放哪个文件？（根据 Guide 职责划分）
• 加了会不会超 token 预算？
• 有没有旧内容该删/归档？

4. 输出建议

📝 [文件名] 该更新了

原因：[什么 trigger 了这次检查]
现在：[当前相关内容]
建议：[具体修改]

要更新吗？

5. 确认后执行

• 用户确认 → 执行修改
• 用户拒绝 → 记录到 memory/ 留底

---

主动检查场景

除了被动 trigger，以下场景主动检查：

1. 新项目讨论完毕 → 检查 MEMORY.md 和 WORKING.md 是否需要加项目
2. 一轮深度对话结束 → 检查有没有新发现该落进 USER.md 或 SOUL.md
3. 新 docs/ 文件创建后 → 检查 AGENTS.md 索引是否需要加入
4. 安装新 Skill 后 → 检查 TOOLS.md 是否需要加规则
5. 用户明显改变沟通风格 → 检查 USER.md 是否需要更新偏好

---

规则

1. 不自动改文件 — 必须用户确认
2. 一次只建议改一个文件
3. 遵守 Guide 的职责划分和 token 预算
4. 检查重复 — 已有类似内容不重复加
5. 该删也提 — 过时内容建议删除或归档
6. 措辞匹配 — 新内容风格和文件现有内容一致

Token 预算

文件	最大
SOUL.md	800B
USER.md	800B
IDENTITY.md	300B
AGENTS.md	1.5KB
TOOLS.md	1KB
MEMORY.md	1.5KB