memory
关于
The memory skill retrieves and utilizes persistent project memories from previous sessions to provide context-aware assistance. Developers should use it when they need to recall past decisions, check established conventions, or understand user preferences. It is triggered by phrases like "remember when" or "what was our decision about" and accesses a structured JSON memory file.
快速安装
Claude Code
推荐/plugin add https://github.com/jeremylongshore/claude-code-plugins-plusgit clone https://github.com/jeremylongshore/claude-code-plugins-plus.git ~/.claude/skills/memory在 Claude Code 中复制并粘贴此命令以安装该技能
技能文档
Prerequisites
Before using this skill, ensure you have:
- Project memory file at
{baseDir}/.memories/project_memory.json - Read permissions for the memory storage location
- Understanding that memories persist across sessions
- Knowledge of slash commands for manual memory management
Instructions
Step 1: Access Project Memories
Retrieve stored memories from previous sessions:
- Locate memory file using Read tool
- Parse JSON structure containing memory entries
- Identify relevant memories based on current context
- Extract applicable decisions, conventions, or preferences
Step 2: Apply Memories to Current Context
Integrate past decisions into current work:
- Use remembered library/tool choices when making similar decisions
- Apply architectural patterns established in prior sessions
- Reference user preferences for coding style or conventions
- Consider past decisions as context for new features
Step 3: Update Memories When Needed
Store new decisions for future reference:
- Add significant architectural choices
- Document tool or library selections with rationale
- Record user preferences and conventions
- Update changed decisions to avoid conflicts
Step 4: Resolve Memory Conflicts
Handle situations where memories conflict with current requests:
- Prioritize current explicit user requests over stored memories
- Flag conflicts for user awareness when appropriate
- Update memories that have become outdated
- Remove memories that are no longer relevant
Output
The skill provides seamless memory-enhanced responses:
Silent Integration
- Memories applied automatically without announcement
- Decisions informed by historical context
- Consistent behavior aligned with past choices
- Natural incorporation of established patterns
Memory Status
When using slash commands:
- List of all stored memories with timestamps
- Confirmation of newly added memories
- Notification of removed or updated memories
- Summary of applicable memories for current task
Error Handling
Common issues and solutions:
Memory File Not Found
- Error: Cannot locate project memory file
- Solution: Initialize new memory file in standard location, prompt user to set up memory persistence
Conflicting Memories
- Error: Multiple memories contradict each other
- Solution: Apply most recent memory, allow current request to override, suggest cleanup
Invalid Memory Format
- Error: Memory file corrupted or improperly formatted
- Solution: Backup existing file, recreate with valid JSON structure, restore recoverable entries
Permission Denied
- Error: Cannot read or write memory file
- Solution: Check file permissions, request necessary access, use alternative storage location
Resources
Memory Management Commands
/remember [text]- Add new memory to manual_memories array/forget [text]- Remove matching memory from storage/memories- Display all currently stored memories
Best Practices
- Apply memories silently without announcing to user
- Current explicit requests always override stored memories
- Store significant decisions that affect future work
- Regularly review and clean up outdated memories
- Use memories as context, not rigid constraints
Memory Categories
- Architecture decisions: Framework choices, design patterns
- Tool selections: Libraries, dependencies, build tools
- Code conventions: Style preferences, naming patterns
- User preferences: Communication style, detail level
- Project constraints: Performance targets, compatibility requirements
Integration Guidelines
- Memory retrieval happens automatically during task analysis
- Memories inform recommendations and implementation choices
- User can override any memory-based decision at any time
- Regular memory updates keep context current and relevant
GitHub 仓库
相关推荐技能
algorithmic-art
元该Skill使用p5.js创建包含种子随机性和交互参数探索的算法艺术,适用于生成艺术、流场或粒子系统等需求。它能自动生成算法哲学文档(.md)和对应的交互式艺术代码(.html/.js),确保作品原创性避免侵权。开发者可通过定义计算美学理念快速获得可交互的艺术实现方案。
subagent-driven-development
开发该Skill用于在当前会话中执行包含独立任务的实施计划,它会为每个任务分派一个全新的子代理并在任务间进行代码审查。这种"全新子代理+任务间审查"的模式既能保障代码质量,又能实现快速迭代。适合需要在当前会话中连续执行独立任务,并希望在每个任务后都有质量把关的开发场景。
executing-plans
设计该Skill用于当开发者提供完整实施计划时,以受控批次方式执行代码实现。它会先审阅计划并提出疑问,然后分批次执行任务(默认每批3个任务),并在批次间暂停等待审查。关键特性包括分批次执行、内置检查点和架构师审查机制,确保复杂系统实现的可控性。
cost-optimization
其他这个Claude Skill帮助开发者优化云成本,通过资源调整、标记策略和预留实例来降低AWS、Azure和GCP的开支。它适用于减少云支出、分析基础设施成本或实施成本治理策略的场景。关键功能包括提供成本可视化、资源规模调整指导和定价模型优化建议。
