code-refactoring-context-restore
恢复代码重构上下文
此技能帮助AI助手在复杂的重构会话中恢复和重建项目上下文，确保在恢复长时间运行的代码改进任务时的连续性和最小信息丢失。
下载技能 ZIP
在 Claude 中上传
前往 设置 → 功能 → 技能 → 上传技能
开启并开始使用
测试它
正在使用“code-refactoring-context-restore”。 How do I restore context for a Ruby on Rails refactoring project?
预期结果:
For Ruby on Rails projects, focus on restoring: (1) Database schema and migrations - critical for understanding model relationships; (2) Route configuration - maps controller structure; (3) Gemfile dependencies - reveals technology decisions; (4) Controller logic - identifies core business logic. Use incremental restoration starting with routes and schema, then expand to controllers and models based on your token budget.
正在使用“code-refactoring-context-restore”。 What is the best approach for semantic context retrieval in large codebases?
预期结果:
Use a tiered retrieval strategy: (1) Embed project structure as top-level context; (2) Use vector similarity search with code embeddings for semantic retrieval; (3) Apply relevance filtering with cosine similarity threshold of 0.75; (4) Rank results by temporal decay and historical decision weight. This approach balances comprehensiveness with token efficiency.
安全审计
安全Static analysis flagged 16 external_commands patterns and 2 weak_crypto patterns. Evaluation reveals these are FALSE POSITIVES: The external_commands detections are markdown code examples (e.g., `context-restore project:ai-assistant --mode full`) used as documentation, not actual shell execution. The weak_crypto flags are references to cryptographic concepts in documentation text, not actual crypto implementation. This is a documentation/guide skill containing only instructional content and example code snippets.
检测到的模式
质量评分
你能构建什么
恢复中断的重构会话
在离开复杂的重构项目数天或数周后返回时，使用此技能重建完整上下文，包括架构决策、已完成的更改和待处理的任务。
将AI接入现有代码库
新团队成员或新的AI会话可以通过语义上下文检索快速了解项目结构、技术决策和当前开发优先级。
跨项目知识转移
通过提取和适配语义上下文向量，将相关架构模式和决策从一个项目转移到另一个项目。
试试这些提示
I need to restore context for my code refactoring project. The project uses [language/framework]. Can you guide me through retrieving semantic context including architectural decisions and recent changes?
My context window is limited. Help me prioritize which context components to restore first. I have a [X] token budget and need to cover [specific areas].
I am working with multiple AI agents on the same refactoring task. How should I structure context so each agent has consistent understanding of project state and decision history?
I have two context snapshots from different sessions that may have conflicting information about the codebase. What strategies can I use to detect and resolve these conflicts?
最佳实践
- 在做出更改之前，始终根据当前代码库状态验证恢复的上下文
- 使用增量上下文加载以在保持准确性的同时保持在令牌预算范围内
- 实施加密签名以进行上下文完整性验证
避免
- 在没有相关性过滤的情况下恢复完整的历史上下文 - 导致令牌溢出
- 忽略上下文排序中的时间衰减 - 导致过时信息获得优先级
- 不根据实时代码库状态验证就假设上下文是最新的