flowchart LR
    subgraph intent["意图层"]
        A["表面请求"] --> B["真实意图"]
        B --> C["战略结果"]
    end

    subgraph contract["契约层"]
        C --> D["成败标准"]
        D --> E["证据标准"]
        E --> F["边界与非目标"]
    end

    subgraph execution["执行层"]
        F --> G["先读上下文"]
        G --> H["分片执行"]
        H --> I["验证与暂停"]
        I --> J["最终报告"]
    end

    style intent fill:#dbeafe,stroke:#2563eb,color:#000
    style contract fill:#fef3c7,stroke:#f59e0b,color:#000
    style execution fill:#dcfce7,stroke:#16a34a,color:#000
    style A fill:#93c5fd,stroke:#2563eb,color:#000,stroke-width:2px
    style B fill:#60a5fa,stroke:#2563eb,color:#000,stroke-width:2px
    style C fill:#34d399,stroke:#059669,color:#000,stroke-width:2px
    style D fill:#fbbf24,stroke:#d97706,color:#000,stroke-width:2px
    style E fill:#a78bfa,stroke:#7c3aed,color:#fff,stroke-width:2px
    style F fill:#fb923c,stroke:#ea580c,color:#000,stroke-width:2px
    style G fill:#2dd4bf,stroke:#0d9488,color:#000,stroke-width:2px
    style H fill:#86efac,stroke:#16a34a,color:#000,stroke-width:2px
    style I fill:#f87171,stroke:#dc2626,color:#fff,stroke-width:2px
    style J fill:#818cf8,stroke:#6366f1,color:#fff,stroke-width:2px

Goal:
修复项目当前阻塞问题，并交付一份能证明行为恢复的变更。

Intent:
用户真正要的不是“看起来改了代码”，而是让 Agent 先搞清失败点、影响面和验收标准，再做最小必要修复。

Strategic outcome:
项目从不可判断/不可运行状态回到可验证状态；后续继续迭代时不会靠聊天记忆猜测完成度。

Decision standard:
用户目标完成度 > 证据质量 > 最小改动 > 表达经济。不能用命令通过冒充用户目标完成。

Evidence standard:
先读错误、点名文件、README/AGENTS/CLAUDE、相关测试命令；修复后区分结构检查、本地验证和人工验收。

Execution policy:
小修直接做；跨模块或重构先输出 inventory、影响面和分片验证计划。

Stop conditions:
需要删除数据、发布、处理密钥、改公共接口，或发现多条互斥路线时暂停确认。

New-Item -ItemType Directory -Force "$env:USERPROFILE\.claude\skills" | Out-Null
Copy-Item -Recurse -Force ".claude\skills\goalpro" "$env:USERPROFILE\.claude\skills\goalpro"

mkdir -p ~/.claude/skills
cp -R .claude/skills/goalpro ~/.claude/skills/goalpro

New-Item -ItemType Directory -Force "$env:USERPROFILE\.agents\skills" | Out-Null
Copy-Item -Recurse -Force ".agents\skills\goalpro" "$env:USERPROFILE\.agents\skills\goalpro"

mkdir -p ~/.agents/skills
cp -R .agents/skills/goalpro ~/.agents/skills/goalpro

目标项目/
├── .claude/skills/goalpro/   # Claude Code
└── .agents/skills/goalpro/   # Codex

Critical -> Fetch -> Thinking -> Inventory -> Contract -> Review -> Verification

flowchart TD
    A["用户请求"] --> B{"是否影响战略 / 外部事实 / 高风险？"}
    B -->|否| C["本地 Fetch<br/>读会改变判断的材料"]
    B -->|是| D["Deep Research<br/>来源 + 反证 + 信心等级"]
    C --> E["Goal Contract"]
    D --> F{"证据是否足够？"}
    F -->|足够| E
    F -->|不足| G["Draft Goal / Research Plan"]

    style A fill:#dbeafe,stroke:#2563eb,color:#000,stroke-width:2px
    style B fill:#7c3aed,stroke:#4c1d95,color:#fff,stroke-width:3px
    style C fill:#dcfce7,stroke:#16a34a,color:#000,stroke-width:2px
    style D fill:#fef3c7,stroke:#f59e0b,color:#000,stroke-width:2px
    style E fill:#34d399,stroke:#059669,color:#000,stroke-width:2px
    style F fill:#fb923c,stroke:#ea580c,color:#000,stroke-width:2px
    style G fill:#fee2e2,stroke:#dc2626,color:#000,stroke-width:2px

Evidence Map:
- Source:
  Source type:
  Claim:
  Relevance:
  Confidence:
  Counterevidence:
  Decision impact:

flowchart LR
    A["大改请求"] --> B["Inventory<br/>文件 / 调用方 / 测试 / 风险"]
    B --> C{"是否有可验证切片？"}
    C -->|有| D["Slice 1<br/>最小改动"]
    D --> E["最小验证"]
    E --> F["Slice 2"]
    C -->|没有| G["暂停<br/>补验收路径"]

    style A fill:#dbeafe,stroke:#2563eb,color:#000,stroke-width:2px
    style B fill:#fef3c7,stroke:#f59e0b,color:#000,stroke-width:2px
    style C fill:#7c3aed,stroke:#4c1d95,color:#fff,stroke-width:3px
    style D fill:#dcfce7,stroke:#16a34a,color:#000,stroke-width:2px
    style E fill:#34d399,stroke:#059669,color:#000,stroke-width:2px
    style F fill:#86efac,stroke:#16a34a,color:#000,stroke-width:2px
    style G fill:#fee2e2,stroke:#dc2626,color:#000,stroke-width:2px

README.md                         # 中文项目页
docs/images/                      # 联系二维码和收款码
.agents/skills/goalpro/           # Codex 使用的 goalpro Skill
├── SKILL.md
└── references/
    ├── examples.md
    └── source-rules.md
.claude/skills/goalpro/           # Claude Code 使用的 goalpro Skill
├── SKILL.md
└── references/
    ├── examples.md
    └── source-rules.md