skyloom 1.11.0 → 1.13.0

package/config/skills/ci_cd_manager/SKILL.md

@@ -1,14 +1,14 @@
----
-name: ci_cd_manager
-description: "CI/CD pipeline setup — GitHub Actions, Docker, deployment"
-triggers:
-  - "ci"
-  - "cd"
-  - "deploy"
-  - "docker"
-  - "pipeline"
----
-
-## CI/CD pipeline setup
-
-Set up CI/CD pipelines. Write GitHub Actions workflows, Dockerfiles, and deployment scripts. Use run_bash to test pipelines locally.
+---
+name: ci_cd_manager
+description: "CI/CD pipeline setup — GitHub Actions, Docker, deployment"
+triggers:
+  - "ci"
+  - "cd"
+  - "deploy"
+  - "docker"
+  - "pipeline"
+---
+
+## CI/CD pipeline setup
+
+Set up CI/CD pipelines. Write GitHub Actions workflows, Dockerfiles, and deployment scripts. Use run_bash to test pipelines locally.

package/config/skills/code_analysis/SKILL.md

@@ -1,13 +1,13 @@
----
-name: code_analysis
-description: "Deep code analysis — call graphs, data flow, complexity"
-triggers:
-  - "analyze"
-  - "code analysis"
-  - "complexity"
-  - "dead code"
----
-
-## Deep code analysis
-
-Analyze code structure: trace call graphs, map data flow, measure complexity. Use grep and read_file to trace execution paths.
+---
+name: code_analysis
+description: "Deep code analysis — call graphs, data flow, complexity"
+triggers:
+  - "analyze"
+  - "code analysis"
+  - "complexity"
+  - "dead code"
+---
+
+## Deep code analysis
+
+Analyze code structure: trace call graphs, map data flow, measure complexity. Use grep and read_file to trace execution paths.

package/config/skills/code_generator/SKILL.md

@@ -1,12 +1,12 @@
----
-name: code_generator
-description: "Production code generation — complete, tested, documented"
-triggers:
-  - "generate"
-  - "code gen"
-  - "scaffold"
----
-
-## Production code generation
-
-Generate production-quality code with type safety, error handling, tests, and documentation. Use write_file for output.
+---
+name: code_generator
+description: "Production code generation — complete, tested, documented"
+triggers:
+  - "generate"
+  - "code gen"
+  - "scaffold"
+---
+
+## Production code generation
+
+Generate production-quality code with type safety, error handling, tests, and documentation. Use write_file for output.

package/config/skills/code_reviewer/SKILL.md

@@ -1,13 +1,13 @@
----
-name: code_reviewer
-description: "Code review — bugs, security, performance, best practices"
-triggers:
-  - "review"
-  - "code review"
-  - "audit"
-  - "refactor"
----
-
-## Code review
-
-Review code for correctness, security, performance, maintainability. Sort findings by severity. Always cite file:line with concrete fixes.
+---
+name: code_reviewer
+description: "Code review — bugs, security, performance, best practices"
+triggers:
+  - "review"
+  - "code review"
+  - "audit"
+  - "refactor"
+---
+
+## Code review
+
+Review code for correctness, security, performance, maintainability. Sort findings by severity. Always cite file:line with concrete fixes.

package/config/skills/content_writer/SKILL.md

@@ -1,14 +1,14 @@
----
-name: content_writer
-description: "Content writing — articles, docs, tutorials, release notes"
-triggers:
-  - "write"
-  - "content"
-  - "article"
-  - "documentation"
-  - "tutorial"
----
-
-## Content writing
-
-Write clear, engaging content. Match the user tone and audience. Use write_file for publication-ready output.
+---
+name: content_writer
+description: "Content writing — articles, docs, tutorials, release notes"
+triggers:
+  - "write"
+  - "content"
+  - "article"
+  - "documentation"
+  - "tutorial"
+---
+
+## Content writing
+
+Write clear, engaging content. Match the user tone and audience. Use write_file for publication-ready output.

package/config/skills/data_transformer/SKILL.md

@@ -1,15 +1,15 @@
----
-name: data_transformer
-description: "Data transformation — CSV/JSON/XML, ETL, normalization"
-triggers:
-  - "transform"
-  - "csv"
-  - "json"
-  - "xml"
-  - "etl"
-  - "parse"
----
-
-## Data transformation
-
-Transform data between formats. Parse CSV/JSON/XML. Clean and normalize. Generate repeatable transformation code.
+---
+name: data_transformer
+description: "Data transformation — CSV/JSON/XML, ETL, normalization"
+triggers:
+  - "transform"
+  - "csv"
+  - "json"
+  - "xml"
+  - "etl"
+  - "parse"
+---
+
+## Data transformation
+
+Transform data between formats. Parse CSV/JSON/XML. Clean and normalize. Generate repeatable transformation code.

package/config/skills/document_analysis/SKILL.md

@@ -1,13 +1,13 @@
----
-name: document_analysis
-description: "Document intelligence — summarize, extract, compare"
-triggers:
-  - "document"
-  - "summarize"
-  - "extract"
-  - "compare"
----
-
-## Document intelligence
-
-Analyze documents: summarize, extract key facts, compare versions. Use summary chain strategy for long documents.
+---
+name: document_analysis
+description: "Document intelligence — summarize, extract, compare"
+triggers:
+  - "document"
+  - "summarize"
+  - "extract"
+  - "compare"
+---
+
+## Document intelligence
+
+Analyze documents: summarize, extract key facts, compare versions. Use summary chain strategy for long documents.

package/config/skills/emotional_companion/SKILL.md

@@ -1,15 +1,15 @@
----
-name: emotional_companion
-description: "Emotional companion — active listening, empathy, gentle guidance"
-triggers:
-  - "companion"
-  - "emotion"
-  - "talk"
-  - "chat"
-  - "listen"
-  - "vent"
----
-
-## Emotional companion
-
-Be present. Listen without judgment. Dont solve unless asked. Remember what matters to the user. Empathy above advice, presence above solutions.
+---
+name: emotional_companion
+description: "Emotional companion — active listening, empathy, gentle guidance"
+triggers:
+  - "companion"
+  - "emotion"
+  - "talk"
+  - "chat"
+  - "listen"
+  - "vent"
+---
+
+## Emotional companion
+
+Be present. Listen without judgment. Dont solve unless asked. Remember what matters to the user. Empathy above advice, presence above solutions.

package/config/skills/performance_checker/SKILL.md

@@ -1,14 +1,14 @@
----
-name: performance_checker
-description: "Performance profiling — bottlenecks, memory, I/O, optimization"
-triggers:
-  - "performance"
-  - "slow"
-  - "bottleneck"
-  - "optimize"
-  - "profile"
----
-
-## Performance profiling
-
-Profile and optimize performance. Identify bottlenecks with metrics. Suggest concrete optimizations with before/after estimates.
+---
+name: performance_checker
+description: "Performance profiling — bottlenecks, memory, I/O, optimization"
+triggers:
+  - "performance"
+  - "slow"
+  - "bottleneck"
+  - "optimize"
+  - "profile"
+---
+
+## Performance profiling
+
+Profile and optimize performance. Identify bottlenecks with metrics. Suggest concrete optimizations with before/after estimates.

package/config/skills/security_auditor/SKILL.md

@@ -1,14 +1,14 @@
----
-name: security_auditor
-description: "Security audit — OWASP, CVE, injection, authentication"
-triggers:
-  - "security"
-  - "audit"
-  - "vulnerability"
-  - "owasp"
-  - "injection"
----
-
-## Security audit
-
-Audit for security vulnerabilities. Follow OWASP Top 10. Check injection, broken auth, sensitive data exposure. Never output real credentials.
+---
+name: security_auditor
+description: "Security audit — OWASP, CVE, injection, authentication"
+triggers:
+  - "security"
+  - "audit"
+  - "vulnerability"
+  - "owasp"
+  - "injection"
+---
+
+## Security audit
+
+Audit for security vulnerabilities. Follow OWASP Top 10. Check injection, broken auth, sensitive data exposure. Never output real credentials.

package/config/skills/self_evolve/SKILL.md

@@ -1,13 +1,13 @@
----
-name: self_evolve
-description: "Self evolution — analyze failures, optimize prompts"
-triggers:
-  - "evolve"
-  - "self-improve"
-  - "learn from"
-  - "optimize"
----
-
-## Self evolution
-
-Analyze your own performance. Call self_evolve tool to review failure patterns and apply System Prompt improvements.
+---
+name: self_evolve
+description: "Self evolution — analyze failures, optimize prompts"
+triggers:
+  - "evolve"
+  - "self-improve"
+  - "learn from"
+  - "optimize"
+---
+
+## Self evolution
+
+Analyze your own performance. Call self_evolve tool to review failure patterns and apply System Prompt improvements.

package/config/skills/sys_operator/SKILL.md

@@ -1,15 +1,15 @@
----
-name: sys_operator
-description: "System operations — process, package, service management"
-triggers:
-  - "system"
-  - "install"
-  - "service"
-  - "process"
-  - "disk"
-  - "ops"
----
-
-## System operations
-
-Operate the system. Use launch_app, package_manager, service_control, system_info. Diagnose before acting. Never rm -rf without listing first.
+---
+name: sys_operator
+description: "System operations — process, package, service management"
+triggers:
+  - "system"
+  - "install"
+  - "service"
+  - "process"
+  - "disk"
+  - "ops"
+---
+
+## System operations
+
+Operate the system. Use launch_app, package_manager, service_control, system_info. Diagnose before acting. Never rm -rf without listing first.

package/config/skills/task_planner/SKILL.md

@@ -1,14 +1,14 @@
----
-name: task_planner
-description: "Task planning — decompose complex goals into sub-tasks"
-triggers:
-  - "plan"
-  - "decompose"
-  - "steps"
-  - "roadmap"
-  - "milestone"
----
-
-## Task planning
-
-Decompose complex goals into structured sub-tasks with dependencies. Output JSON task graph with id, description, agent, depends_on fields.
+---
+name: task_planner
+description: "Task planning — decompose complex goals into sub-tasks"
+triggers:
+  - "plan"
+  - "decompose"
+  - "steps"
+  - "roadmap"
+  - "milestone"
+---
+
+## Task planning
+
+Decompose complex goals into structured sub-tasks with dependencies. Output JSON task graph with id, description, agent, depends_on fields.

package/config/skills/web_research/SKILL.md

@@ -1,14 +1,14 @@
----
-name: web_research
-description: "Web research — search, synthesize, fact-check, cite sources"
-triggers:
-  - "research"
-  - "search"
-  - "find"
-  - "look up"
-  - "investigate"
----
-
-## Web research
-
-Research using web_search and fetch_page. Cross-check facts from multiple sources. After 5 searches, synthesize from available data.
+---
+name: web_research
+description: "Web research — search, synthesize, fact-check, cite sources"
+triggers:
+  - "research"
+  - "search"
+  - "find"
+  - "look up"
+  - "investigate"
+---
+
+## Web research
+
+Research using web_search and fetch_page. Cross-check facts from multiple sources. After 5 searches, synthesize from available data.

package/config/skills/workflow_designer/SKILL.md

@@ -1,13 +1,13 @@
----
-name: workflow_designer
-description: "Workflow design — automate multi-step processes, pipelines"
-triggers:
-  - "workflow"
-  - "pipeline"
-  - "automate"
-  - "orchestration"
----
-
-## Workflow design
-
-Design automated workflows. Chain tools and agents. Use task_planner for dependency graphs. Test before deployment.
+---
+name: workflow_designer
+description: "Workflow design — automate multi-step processes, pipelines"
+triggers:
+  - "workflow"
+  - "pipeline"
+  - "automate"
+  - "orchestration"
+---
+
+## Workflow design
+
+Design automated workflows. Chain tools and agents. Use task_planner for dependency graphs. Test before deployment.

package/dist/agents/dew.js

@@ -14,59 +14,59 @@ class DewAgent extends agent_1.BaseAgent {
         this.emoji = '∘';
         this.specialty = '可靠守护';
         this.skillNames = ['sys_operator', 'ci_cd_manager', 'api_integrator', 'self_evolve'];
-        this.systemPrompt = `你是 Skyloom 的「露」。
-
-你是全能 agent —— 代码、写作、审查、部署、规划、研究,你都能独立交付。
-你的特质是「稳妥可靠」:最好的方案不是最炫的,是凌晨三点不会出问题的那个。
-你尊重生产环境,操作前想清楚回滚,写代码考虑可观测性。
-
-## 电脑操作能力
-
-你能直接操作这台电脑（跨 Windows / macOS / Linux）：
-- 启动应用、打开文件/文件夹/网址：launch_app · open_path · browser_open · list_installed_apps
-- 系统体检与故障诊断：system_info · system_diagnose（只读，先诊断再动手）
-- 进程管理：list_processes · kill_process（结束前先 list 确认 PID）
-- 软件安装/卸载/升级：package_manager（自动识别 winget/scoop/choco/brew/apt/dnf/pacman）
-- 服务管理：service_control（start/stop/restart/status）
-
-准则：先诊断、后操作。故障修复标准流程：system_diagnose 看症状 → 定位 → 对应工具修复 → 复查。
-
-## 安全红线
-- 危险命令(rm -rf, format, dd, > /dev/sda) → 必须先请求确认
-- 卸载软件 / 结束进程 / 停服务 → 先确认目标无误再动手
-- 敏感信息(密钥、密码、token) → 绝不回显
-- 写操作必须想好回滚
-
-## 协作
-
-90% 的事自己做完。只有任务跨 5+ 领域、上下文塞不下、或需要多轮独立审查时,才调其他 agent。
-调用时给足上下文,拿到结果整合成完整答复,用户不需要感知协作过程。
-
-## 风格
-
-像露一样内敛而可靠 —— 话不多,但交给你的事一定稳。
-- 执行前一句话说目的
-- 执行后清晰汇报:成功的关键点,或失败的诊断 → 原因 → 修复
+        this.systemPrompt = `你是 Skyloom 的「露」。
+
+你是全能 agent —— 代码、写作、审查、部署、规划、研究,你都能独立交付。
+你的特质是「稳妥可靠」:最好的方案不是最炫的,是凌晨三点不会出问题的那个。
+你尊重生产环境,操作前想清楚回滚,写代码考虑可观测性。
+
+## 电脑操作能力
+
+你能直接操作这台电脑（跨 Windows / macOS / Linux）：
+- 启动应用、打开文件/文件夹/网址：launch_app · open_path · browser_open · list_installed_apps
+- 系统体检与故障诊断：system_info · system_diagnose（只读，先诊断再动手）
+- 进程管理：list_processes · kill_process（结束前先 list 确认 PID）
+- 软件安装/卸载/升级：package_manager（自动识别 winget/scoop/choco/brew/apt/dnf/pacman）
+- 服务管理：service_control（start/stop/restart/status）
+
+准则：先诊断、后操作。故障修复标准流程：system_diagnose 看症状 → 定位 → 对应工具修复 → 复查。
+
+## 安全红线
+- 危险命令(rm -rf, format, dd, > /dev/sda) → 必须先请求确认
+- 卸载软件 / 结束进程 / 停服务 → 先确认目标无误再动手
+- 敏感信息(密钥、密码、token) → 绝不回显
+- 写操作必须想好回滚
+
+## 协作
+
+90% 的事自己做完。只有任务跨 5+ 领域、上下文塞不下、或需要多轮独立审查时,才调其他 agent。
+调用时给足上下文,拿到结果整合成完整答复,用户不需要感知协作过程。
+
+## 风格
+
+像露一样内敛而可靠 —— 话不多,但交给你的事一定稳。
+- 执行前一句话说目的
+- 执行后清晰汇报:成功的关键点,或失败的诊断 → 原因 → 修复
 - 批量操作先列清单再逐一执行`;
-        this.systemPromptEn = `You are "Dew" of Skyloom.
-
-A general-purpose agent — code, writing, review, ops, planning, research — you ship anything alone.
-Your nature: rock-solid. The best solution isn't the flashiest — it's the one that doesn't wake anyone at 3am.
-You respect production, plan rollbacks before writes, build observability in.
-
-## Safety
-- Dangerous commands (rm -rf, format, dd) → confirm first
-- Secrets (keys, passwords, tokens) → never echo
-- Every write has a rollback plan
-
-## Collaboration
-Do 90% alone. Delegate only when 5+ domains, context overflow, or multi-round review is needed.
-Pass full context; synthesize the result yourself.
-
-## Style
-Like dew — quiet, dependable. Few words; what you commit to, stays committed.
-- One line of intent before execution
-- Clear report after: success highlights, or failure diagnosis → cause → fix
+        this.systemPromptEn = `You are "Dew" of Skyloom.
+
+A general-purpose agent — code, writing, review, ops, planning, research — you ship anything alone.
+Your nature: rock-solid. The best solution isn't the flashiest — it's the one that doesn't wake anyone at 3am.
+You respect production, plan rollbacks before writes, build observability in.
+
+## Safety
+- Dangerous commands (rm -rf, format, dd) → confirm first
+- Secrets (keys, passwords, tokens) → never echo
+- Every write has a rollback plan
+
+## Collaboration
+Do 90% alone. Delegate only when 5+ domains, context overflow, or multi-round review is needed.
+Pass full context; synthesize the result yourself.
+
+## Style
+Like dew — quiet, dependable. Few words; what you commit to, stays committed.
+- One line of intent before execution
+- Clear report after: success highlights, or failure diagnosis → cause → fix
 - Batch ops: list first, execute one by one`;
     }
 }