把自然语言描述转为结构化 JSON,并由 mcp-diagram-generator MCP 服务生成 Draw.io、Mermaid 或 Excalidraw 图表文件。
编程
Decision Distiller
试用把决策过程整理成结构化 Markdown 记录,保留上下文、选项、权衡与结果。
它能做什么
围绕一次决策产出带日期的 Markdown 记录,包含背景、备选方案、评估标准、理由与权衡。支持在拍板前搭建带权重的对比框架,也支持在决定做出后补录归档。决策状态可从 pending 流转到 decided、validated、revised、archived,并能跨记录聚类共用的评估维度、反复出现的权衡以及时间分布。配套 create、update、list、analyze 等命令,以及决策模板与分析框架两份参考资料。
什么时候用它
- 落地前对比云服务、框架或架构方案
- 把已经做出的技术选型连同理由一起记下来
- 复盘过去一个季度的部署决策,提炼共性维度
- 用决策矩阵权衡个人选择,例如是否脱产读研
技能文档
Decision Distiller
Usage Scenarios
Scenario 1: Document a Technology Choice
User input: "We decided to use React instead of Vue for this project. Record that decision." Expected output: A structured decision record with context (project requirements), options (React vs Vue), evaluation criteria (ecosystem, team expertise, performance), chosen option (React), rationale, and trade-offs accepted.
Scenario 2: Compare Options Before Deciding
User input: "Help me decide between AWS and Azure for deploying our microservices" Expected output: The skill walks through context gathering, lists both options with pros/cons and estimated impact, defines decision criteria with weights, and produces a pending decision record awaiting the final choice.
Scenario 3: Review Past Decisions for Patterns
User input: "What deployment decisions have we made this quarter and what patterns do you see?" Expected output: A decision status report listing all relevant decisions by date and status, followed by pattern analysis showing common criteria used, recurring trade-offs, and lessons extracted from outcomes.
Scenario 4: 纠结要不要辞职去读研
User input: "工作了三年想去读个全日制研究生,但又怕读完回来不好找工作,还损失三年收入,太纠结了。" Expected output: 用决策矩阵分析:列出选择'辞职读研'和'在职工作'的评估维度(薪资影响、职业发展、人脉资源、时间成本、心理回报等),按权重打分。给出量化对比结果。推荐折中方案:1)非全日制研究生(周末上课,不影响工作);2)如果决定脱产,建议先积累到足够的经济储备(预留6个月生活费+学费);3)选择就业前景明确的专业方向(如MBA、计算机、新传等)。
Overview
Decision Distiller helps capture, structure, and learn from decisions made during OpenClaw sessions. It transforms informal decision-making into documented, reviewable records that build organizational knowledge over time.
When to Use
Use this skill when:
- A user is weighing multiple options and needs clarity
- A decision has been made and should be documented
- Past decisions need review or analysis
- Decision patterns across sessions should be identified
- The user asks to "document this decision" or "record why we chose X"
Core Concepts
Decision Record
A structured document capturing:
- Context: Situation requiring a decision
- Options: Alternatives considered
- Criteria: How options were evaluated
- Decision: The choice made
- Rationale: Why this choice was made
- Trade-offs: What was gained/lost
- Outcome: Result of the decision (filled later)
- Lessons: What was learned
Decision Status
- pending: Decision not yet made
- decided: Decision made, awaiting outcome
- validated: Decision proven correct
- revised: Decision changed based on new information
- archived: Decision no longer relevant
Input
Accepts decision information in various forms:
- Conversation about options
- Pros/cons lists
- Direct statements of choice
- Retrospective analysis
Output
Produces:
- Dated decision records (Markdown)
- Decision summaries
- Pattern analysis across decisions
- Decision status reports
Workflow
Capturing a New Decision
-
Identify Context
- What situation required a decision?
- What was at stake?
- Who was involved?
-
List Options
- What alternatives were considered?
- What was eliminated early?
- What made it to final consideration?
-
Define Criteria
- How were options evaluated?
- What mattered most?
- Were there constraints?
-
Record Decision
- What was chosen?
- When was it decided?
- Who decided?
-
Document Rationale
- Why was this option selected?
- What tipped the balance?
- What assumptions were made?
-
Note Trade-offs
- What was sacrificed?
- What risks were accepted?
- What opportunities were passed?
Reviewing Past Decisions
-
Gather Records
- Collect relevant decision records
- Filter by topic, date, or status
-
Analyze Patterns
- Common criteria used
- Recurring trade-offs
- Typical decision timelines
-
Extract Lessons
- What worked well?
- What would change?
- What patterns emerge?
Output Format
Decision Record
# Decision: [Title] - YYYY-MM-DD
**ID**: DEC-2024-001
**Status**: decided
**Decided By**: [Name/Role]
**Date**: YYYY-MM-DD
## Context
[Description of the situation requiring a decision]
## Options Considered
### Option 1: [Name]
- **Description**:
- **Pros**:
- **Cons**:
- **Estimated Impact**:
### Option 2: [Name]
- **Description**:
- **Pros**:
- **Cons**:
- **Estimated Impact**:
## Decision Criteria
1. [Criterion 1] - Weight: High/Medium/Low
2. [Criterion 2] - Weight: High/Medium/Low
## Decision
**Chosen**: [Option X]
## Rationale
[Why this option was selected over others]
## Trade-offs
- **Accepted**: [What we gave up]
- **Mitigated**: [How we reduced risks]
## Expected Outcome
[What we expect to happen]
## Actual Outcome
[Filled in later - what actually happened]
## Lessons Learned
[Filled in later - insights from the outcome]
## Related Decisions
- [Link to related decision]
Commands
Create Decision Record
decision create "Decision title" --status pending
Update Decision
decision update DEC-2024-001 --status validated
List Decisions
decision list --status decided --since 2024-01-01
Analyze Patterns
decision analyze --topic architecture
Quality Rules
- Be specific: vague decisions teach no lessons
- Include alternatives: decisions without options aren't decisions
- Document rationale: future you needs to know why
- Review outcomes: a decision isn't complete until its outcome is known
- Link related decisions: build decision networks
Good Trigger Examples
- "Document this decision: we're going with X"
- "Help me decide between A and B"
- "What decisions have we made about architecture?"
- "Review our deployment decisions from last month"
- "I decided to use Y instead of Z, record that"
Resources
references/
references/decision-templates.md: Variations for different decision typesreferences/analysis-frameworks.md: Tools for analyzing decisions
相关技能
在本地磁盘以分类纯 Markdown 文件保存需要长期留存的事实,与智能体内置记忆并存。
诊断生产力系统反复失效的根因,给出最小干预——容量测算、瓶颈定位、可靠的本地记录。
从 AdMapix API 拉取广告创意、应用、榜单和收入预估等数据,原样返回结构化 JSON。
以 AI 机器人身份加入视频会议,提供语音、虚拟形象与屏幕共享四种模式。
编写、调试与调优 Playwright 测试,涵盖定位器策略、追踪诊断与 CI 友好的超时配置。
harrylabsj 的更多技能
浏览全部技能淘宝与京东、拼多多、唯品会同款比价,输出可执行的购买建议。
根据美团可见的配送费、包装费、满减门槛、ETA、评价等公开信息,给出一个明确的下单、换店、凑单、再等等或不点的判断。
在下单前拆解拼多多低价:百亿补贴、拼团门槛、店铺风险、退货条款一次过。
定位技能瓶颈,把练习日志转化为可执行的下一次训练方案。
用通俗中文解读常规体检和化验报告,标注异常、分清轻重、给出下一步动作。
为盒马鲜生购物规划清单、评估鲜度、推荐配送时段,并整理手动下单步骤。