file-history
在查找特定文件的过去工作、了解编写代码的原因或用户在文件/功能上引用“我们之前所做的事情”时使用。使用文件历史记录跟踪以获得准确的结果。对于特定于文件的查询,优于情景内存。
📖 中文导读:在查找特定文件的过去工作、了解编写代码的原因或用户在文件/功能上引用“我们之前所做的事情”时使用。使用文件历史记录跟踪以获得准确的结果。对于特定于文件的查询,优于情景内存。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: file-history description: Use when looking for past work on specific FILES, understanding why code was written, or when user references "what we did before" on a file/feature. Uses file-history tracking for accurate results. PREFER THIS over episodic-memory for file-specific queries.
File History Context
Use this for FILE-SPECIFIC history queries. For general semantic search across all conversations, use episodic-memory instead.
When to Use This Skill
USE file-history when:
- Looking for sessions that edited a specific file
- Understanding why code was written a certain way
- User says "what did we do with [filename]"
- User asks about past changes to a component/module
- Refactoring and need to understand original intent
- Code seems unusual and you want the backstory
DON'T use this when:
- General conceptual questions across all projects → use episodic-memory
- No specific file/feature in mind → use episodic-memory
- Looking for patterns across unrelated conversations → use episodic-memory
How to Use
Dispatch the history-context agent:
Task tool:
description: "Find history for [file/feature]"
prompt: "Find context for [specific file or feature]. Focus on [decisions/implementation/rationale]."
subagent_type: "history-context"
Example prompts:
- "Find context for src/components/AuthForm.tsx"
- "What sessions worked on the caching implementation?"
- "Get history for the payment processing feature"
What You Get Back
The agent will return:
- Summary of what was done and why
- Key decisions and rationale
- Implementation notes
- Full session IDs (UUIDs) for reference
- Session file paths for direct access
Why This Works Better Than episodic-memory for Files
| Aspect | file-history | episodic-memory |
|---|---|---|
| Accuracy | Uses Claude's file-history tracking | Semantic search (fuzzy) |
| Speed | Indexed, instant | Searches full conversations |
| File mapping | Knows exactly which sessions edited which files | Guesses from content |
| Git dependency | None | None |
| Project scope | Per-project | Cross-project |
Technical Details
Uses ~/.claude/commands/claude-history.py which:
- Scans
~/.claude/file-history/to build file→session index - Cross-references with
~/.claude/projects/for session metadata - Reads and synthesizes conversation content
- Returns structured context with full session IDs