数据库生成
- 作者仓库星标 34
- 作者更新于 实时读取
- 作者仓库 dotfiles
- 领域
- 工程开发
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @steveclarke · 未声明 license
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 即装即用
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- macOS · Linux
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。
---
name: handoff
description: Prepare a clean handoff for continuing complex work in a fresh Claude Code session. Use on 'hand…
category: 工程开发
runtime: 无特殊运行时
---
# handoff 输出预览
## PART A: 任务判断
- 适用问题:代码实现、重构、调试或代码审查。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“Why This Matters / Process / 1. Gather State”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于代码实现、重构、调试或代码审查,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“Why This Matters / Process / 1. Gather State”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、写入/修改文件、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文没有稳定的斜杠命令要求。安装验证后通常全局生效,直接在对话里点名这个 Skill 并描述任务即可。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、写入/修改文件。
先用一个小任务确认它会围绕“Why This Matters / Process / 1. Gather State”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: handoff
description: Prepare a clean handoff for continuing complex work in a fresh Claude Code session. Use on 'hand…
category: 工程开发
source: steveclarke/dotfiles
---
# handoff
## 什么时候使用
- handoff 是一个工程开发方向的技能,扩展 Agent 在写代码、做 review、跑测试这类场景下的能力 适合处理工程开发场景下的代码实现、调试、重构、测试或代码审查,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 A…
- 面向代码实现、重构、调试或代码审查,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「Why This Matters / Process / 1. Gather State」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件;主要在本地完成;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "handoff" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Why This Matters / Process / 1. Gather State
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} Handoff — Context-to-Context Continuation
Write durable continuation notes to the project, generate a resume prompt, and copy it to the clipboard so a fresh Claude Code session can pick up exactly where this one left off.
Why This Matters
Claude Code sessions have finite context. Complex multi-session work (migrations, large features, multi-phase plans) needs a way to pass state between sessions without losing momentum. The handoff captures what a fresh session needs to know — not everything that happened, but everything that matters going forward.
Process
1. Gather State
Before writing anything, collect this information from the current session:
What's done:
- Tasks/phases completed
- Files created or modified (high-level, not every file)
- Commits made and pushed (branch name, remote status)
What's not done:
- Remaining tasks or phases
- Known issues, bugs encountered and fixed (so the next session doesn't repeat them)
- Anything partially started
Key decisions made:
- Architectural choices the user approved
- Patterns established that future work should follow
- Things that were tried and didn't work
Practical details:
- Repo location, branch, remote
- Dev environment state (running? needs setup?)
- Test credentials if relevant
- Relevant file paths (specs, plans, docs)
If the next task isn't obvious from context, ask: "What should the next session focus on?"
2. Write Continuation File
Write a continuation notes file to the project. Location preference:
docs/superpowers/plans/if it exists (superpowers convention)docs/if it exists- Project root as
CONTINUATION.md
Filename: YYYY-MM-DD-<topic>-continuation.md
Structure:
# [Topic] — Continuation Notes
## Current State (YYYY-MM-DD)
[1-2 sentence summary of where things stand]
### What's Working
- [bullet list of completed work]
### What's NOT Working Yet
- [bullet list of remaining work, known gaps]
### Key Decisions
- [decisions the user approved that constrain future work]
### Bugs Fixed (Don't Repeat These)
- [gotchas encountered — saves the next session from re-discovering them]
### Practical Details
- Repo: [path] (GitHub: [org/repo], branch: [branch])
- Dev stack: [running? how to start?]
- Credentials: [if needed for testing]
## Next: [What To Do]
[Brief description of what the next session should do first]
### Key Files to Read
- [file path] — [what it contains and why it matters]
### Pattern to Follow
[If a pattern was established, show a brief example so the next session
doesn't have to rediscover it]
Keep it concise — this is a briefing, not a history. The next session should be able to read this in under a minute and know exactly what to do.
3. Commit and Push
git add <continuation-file>
git commit -m "docs: add continuation notes for handoff"
git push
4. Generate Resume Prompt
Craft a resume prompt — the exact text the user will paste into a fresh session. It should:
- Be self-contained (no "see above" references)
- Point to 2-4 files max (the continuation notes + key docs)
- State the next task clearly
- Be under 10 lines
Template:
I'm [brief context]. [Phase/step] is complete. Read these files to get up to speed, then [next action]:
1. [continuation file path] (where we left off)
2. [spec or plan path] (full context)
3. [CLAUDE.md or other key doc] (conventions)
[One sentence about what to do next.]
5. Copy to Clipboard
echo "<resume prompt>" | pbcopy # macOS
# or: echo "<resume prompt>" | xclip -selection clipboard # Linux
Tell the user: "Resume prompt copied to clipboard. Start a fresh session in [repo path] and paste it."
What NOT to Include
- Full conversation history or play-by-play
- Raw error logs (summarize the fix, not the stack trace)
- File contents that can be read from disk
- Implementation details that are in the code
- Anything the fresh session can derive by reading the codebase
The continuation file supplements git log and the code itself — it captures context that isn't in either place.
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核