Reflexion Memorize
技能 已验证 活跃Curates insights from reflections and critiques into CLAUDE.md using Agentic Context Engineering
To systematically capture, curate, and organize learnings from AI agent interactions into a durable knowledge base (CLAUDE.md), thereby enhancing future agent performance and consistency.
功能
- Memory consolidation from reflections and critiques
- Updates CLAUDE.md with actionable insights
- Uses Agentic Context Engineering principles
- Improves future agent performance through structured knowledge
使用场景
- Memorizing key learnings after a complex task or reflection session
- Refining agent behavior by storing successful patterns and avoiding anti-patterns
- Building a project-specific knowledge base to guide future AI interactions
- Ensuring consistency and quality in AI-generated outputs over time
非目标
- Generating initial solutions or approaches
- Performing initial reflection or critique
- Replacing the core AI conversation or task execution
- Storing arbitrary or irrelevant information without curation
工作流
- Harvest context from recent reflections and work
- Extract and categorize key insights (grow)
- Analyze current CLAUDE.md context (refine)
- Apply curation rules and transformations
- Update CLAUDE.md with curated insights
- Validate memory updates against quality gates
实践
- Knowledge Management
- Agentic Context Engineering
- Documentation Curation
先决条件
- Claude Code environment
- Access to CLAUDE.md file
安装
请先添加 Marketplace
/plugin marketplace add NeoLabHQ/context-engineering-kit/plugin install reflexion@context-engineering-kit质量评分
已验证类似扩展
ARA Research Manager
100Records research provenance as a post-task epilogue, scanning conversation history at the end of a coding or research session to extract decisions, experiments, dead ends, claims, heuristics, and pivots, and writing them into the ara/ directory with user-vs-AI provenance tags. Use as a session epilogue — never during execution — to maintain a faithful, auditable trace of how a research project actually evolved.
Orchestrate
100Wire Commands, Agents, and Skills together for complex features. Use when building features that need research, planning, and implementation phases.
Agent Teams
100Coordinate multiple Claude Code sessions as a team — lead + teammates with shared task lists, mailbox messaging, and file-lock claiming. Patterns for team sizing, task decomposition, and when to use teams vs sub-agents vs worktrees.
Domain Extract
100Extract domain knowledge from existing project sources and generate domain rules. Also handles vault sync and domain listing.
Lark Wiki CLI
100飞书知识库：管理知识空间、空间成员和文档节点。创建和查询知识空间、查看和管理空间成员、管理节点层级结构、在知识库中组织文档和快捷方式。当用户需要在知识库中查找或创建文档、浏览知识空间结构、查看或管理空间成员、移动或复制节点时使用。
Learn
100捕获一条经验教训作为结构化的知识条目。当您发现一个不明显的 bug、解决方法或值得为未来会话保存的模式时使用。当用户说“记住这个”、“保存这个经验教训”或“这个值得注意”时使用。