在回答前检索过往对话,找回真实上下文并避免重复询问或误判新话题。
复制安装指令,让 AI 自动完成配置 · 推荐新手
请帮我安装 askskill 上的 "remembering-conversations" 技能: 1. 下载 https://raw.githubusercontent.com/obra/episodic-memory/main/skills/remembering-conversations/SKILL.md 2. 保存为 ~/.claude/skills/remembering-conversations/SKILL.md 3. 装好后重载技能,告诉我可以用了
请先检索我们之前关于这个项目的对话,整理已经确认的需求、未决问题和我明确否决过的方案,再继续给出下一步建议。
一份基于历史对话的需求回顾,包含已确认事项、待确认项和后续建议。
在回答这个问题前,先搜索我们过去有没有讨论过相同主题,尤其是我之前做过的决定、限制条件和偏好,然后基于这些信息回答。
结合过往对话的回答,明确引用已有决定与限制,减少重复沟通。
先检索我们以前是否聊过这个问题。如果聊过,请总结上次结论、后来是否有更新,以及这次应从哪里继续,而不是从头开始解释。
对历史讨论的延续性总结,帮助直接接上次进度推进。
Core principle: Search before reinventing. Searching costs nothing; reinventing or repeating mistakes costs everything.
YOU MUST search historical memory for any historical search.
Announce: "Searching past conversations for [topic]."
Use the Task tool with subagent_type: "search-conversations":
Task tool:
description: "Search past conversations for [topic]"
prompt: "Search for [specific query or topic]. Focus on [what you're looking for - e.g., decisions, patterns, gotchas, code examples]."
subagent_type: "search-conversations"
If a search-conversations agent is available, dispatch it with the same prompt. If not, use the MCP tools directly:
search toolread toolThe search workflow will:
search toolread toolSaves 50-100x context vs. loading raw conversations.
Use this whenever the current task would benefit from information you may have learned before, even if the user did not explicitly ask you to search.
When past experience may help:
When you're stuck:
When historical signals are present:
Before answering from uncertainty:
Don't search first:
Use these directly when a search agent is unavailable or the current harness does not support agent dispatch:
mcp__plugin_episodic-memory_episodic-memory__searchmcp__plugin_episodic-memory_episodic-memory__readWhen using MCP tools directly, keep context small: search first, then read only the top 2-5 relevant conversations or line ranges.
See MCP-TOOLS.md for complete API reference if needed for advanced usage.
帮助开发者用早返回或表驱动方式简化嵌套条件分支,提升代码可读性。
为 Claude Code 提供跨会话持久记忆、检索上下文并自动回注信息。