write-continue-here
关于
This Claude Skill generates a structured CONTINUE_HERE.md file to preserve session context for seamless handoffs between coding sessions. It captures objectives, completed work, in-progress tasks, next steps, and relevant context that git alone doesn't track. Use it when ending unfinished work sessions or transferring task state between Claude instances.
快速安装
Claude Code
推荐npx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/write-continue-here在 Claude Code 中复制并粘贴此命令以安装该技能
技能文档
Write Continue Here
Write structured continuation file → next sess starts w/ full ctx.
Use When
- Ending sess w/ work in progress
- Hand off complex task across sess
- Preserve intent, failed approaches, next steps git can't capture
- Pre-close Claude Code mid-task
In
- Required: Active sess w/ recent work to summarize
- Optional: Specific instructions on what to emphasize
Do
Step 1: Assess State
Gather facts:
git log --oneline -5
git status
git diff --stat
Review conv ctx: objective, done, partial, tried+failed, decisions.
Got: Clear task state — done, in-progress, next steps.
If err: Not in repo → skip git. File still captures conv ctx + task state.
Step 2: Write CONTINUE_HERE.md
Write to project root w/ structure below. Every section actionable, no placeholders.
# Continue Here
> Last updated: YYYY-MM-DDTHH:MM:SSZ | Branch: current-branch-name
## Objective
One-paragraph description of what we are trying to accomplish and why.
## Completed
- [x] Finished item with key file paths (e.g., `src/feature.R`)
- [x] Decisions made and their rationale
## In Progress
- [ ] Partially complete work — describe current state (branch, file:line)
- [ ] Known issues with partial work
## Next Steps
1. Immediate next action (most important)
2. Subsequent actions in priority order
3. **[USER]** Items needing user input or decision
## Context
- Failed approaches and why they did not work
- Key constraints or trade-offs discovered
- Relevant issue/PR links
Guidelines:
- Objective: WHY → git log shows what changed, not why
- Completed: Mark clearly done → prevent re-work
- In Progress: Highest-value section → partial state hardest to reconstruct
- Next Steps: Number by priority. User-dep → prefix
**[USER]** - Context: Negative space → tried+rejected, why
Got: CONTINUE_HERE.md at project root, all 5 sections w/ real content. Timestamp + branch accurate.
If err: Write fails → check perms. Create in project root (same dir as .git/). Verify .gitignore has CONTINUE_HERE.md → if not, add.
Step 3: Verify
Read back CONTINUE_HERE.md, confirm:
- Timestamp current (last few min)
- Branch matches
git branch --show-current - All 5 sections real content (no placeholders)
- Next Steps numbered + actionable
- In Progress items specific enough to resume
Got: Reads as clear actionable handoff → fresh sess can resume.
If err: Edit placeholder/vague sections. Each must pass: "Could fresh sess act w/o clarifying?"
Check
- CONTINUE_HERE.md at project root
- All 5 sections real content (no placeholders)
- Timestamp + branch accurate
-
.gitignorehasCONTINUE_HERE.md - Next Steps numbered + actionable
- In Progress items specific enough to resume
Traps
- Placeholders not content: "TODO: fill later" defeats purpose. Real info from current sess.
- Duplicate git state: Don't list every file → git tracks. Focus intent, partial, next.
- Forget Context: Failed approaches most valuable. W/o them, next sess retries dead ends.
- Overwrite w/o read: Existing CONTINUE_HERE.md → read first → may have unfinished from prior.
- Stale files: Ephemeral → after consume, delete. Stale → confusion.
→
read-continue-here— complement: read+act on continuation file at sess startbootstrap-agent-identity— cold-start identity reconstruction → consumes file this producesmanage-memory— durable cross-sess knowledge (complements ephemeral handoff)commit-changes— save to git before writing continuationwrite-claude-md— project instructions where optional continuity guidance lives
GitHub 仓库
相关推荐技能
railway-docs
文档Railway Docs Skill可实时获取最新的Railway官方文档,确保回答的准确性。当开发者询问Railway功能特性、工作原理或分享docs.railway.com链接时,应优先使用此技能。它通过专门的LLM优化文档源提供最新信息,避免依赖过时记忆来回答技术问题。
n8n-code-python
文档该Skill为在n8n平台的Python代码节点中编写代码提供专家指导,特别适用于需要使用_input/_json/_node语法、Python标准库或了解n8n中Python限制的场景。它强调JavaScript应作为首选方案,仅当需要特定Python功能或对Python语法更熟悉时才使用Python。Skill提供了快速入门模板和关键注意事项,帮助开发者在n8n中高效编写Python代码。
archon
文档Archon Skill为开发者提供了基于RAG的语义搜索和项目任务管理功能,可通过REST API访问知识库。它支持文档搜索、网站爬取、文件上传和版本控制,适用于技术文档查询和项目管理场景。首次使用时需要配置Archon主机地址,建议在处理外部文档时优先使用该Skill。
n8n-code-javascript
文档这个Skill为n8n工作流中的JavaScript代码节点提供专业指导,涵盖数据处理、HTTP请求和日期操作等核心场景。它详细解释了如何正确使用n8n特有的`$input`/`$json`语法、`$helpers`工具以及DateTime对象,并包含关键的错误排查和模式选择建议。开发者通过该Skill能快速掌握Code节点的正确返回格式、数据访问方法和常见陷阱解决方案。
