スキル agent-memory-mcp
📦

agent-memory-mcp

安全

Enable Persistent Memory for AI Agents

This skill provides AI agents with persistent, searchable memory to retain architecture decisions, patterns, and knowledge across sessions without manual tracking.

対応: Claude Codex Code(CC)
🥉 74 ブロンズ
1

スキルZIPをダウンロード

2

Claudeでアップロード

設定 → 機能 → スキル → スキルをアップロードへ移動

3

オンにして利用開始

テストする

「agent-memory-mcp」を使用しています。 Search for all stored patterns related to authentication

期待される結果:

Found 3 memories:
- auth-oauth2: OAuth2 implementation pattern
- auth-jwt: JWT token handling pattern
- auth-sessions: Session management pattern

「agent-memory-mcp」を使用しています。 Write a new architecture decision about database choice

期待される結果:

Memory saved successfully:
Key: db-choice-postgres
Type: decision
Content: Selected PostgreSQL for its relational integrity and JSON support
Tags: database, architecture, production

「agent-memory-mcp」を使用しています。 Show memory statistics

期待される結果:

Memory Stats:
- Total memories: 24
- Types: decisions(8), patterns(12), knowledge(4)
- Top tags: authentication(5), database(4), api(3)

セキュリティ監査

安全
v1 • 2/24/2026

All static findings are false positives. The skill contains only documentation with setup instructions (git clone, npm commands). The flagged 'external_commands' are legitimate bash commands in documentation, not executable code. The flagged 'network' URLs are documentation references (GitHub repo URL, localhost dashboard). The flagged 'cryptographic' terms ('key', 'risk') are common English words used in their normal sense, not cryptographic algorithms.

1
スキャンされたファイル
88
解析された行数
0
検出結果
1
総監査数
セキュリティ問題は見つかりませんでした
監査者: claude

品質スコア

38
アーキテクチャ
100
保守性
87
コンテンツ
50
コミュニティ
100
セキュリティ
91
仕様準拠

作れるもの

Retain Architecture Decisions

Store and retrieve architectural choices made during project setup so future sessions understand why certain patterns were selected.

Build Pattern Library

Collect and search reusable code patterns, snippets, and best practices discovered during development.

Track Project Context

Maintain persistent context about project goals, constraints, and conventions across agent sessions.

これらのプロンプトを試す

Save Architecture Decision
Use memory_write to store this architecture decision: key: '{decision_key}', type: 'decision', content: '{decision_content}', tags: ['{tag1}', '{tag2}']
Find Authentication Patterns
Search the memory for authentication patterns using memory_search with query: 'authentication', type: 'pattern'
Recall Design Discussion
Retrieve the design discussion about '{topic}' using memory_read with key: '{memory_key}'
Review Memory Statistics
Show memory statistics using memory_stats to understand what knowledge has been stored

ベストプラクティス

  • Use consistent naming conventions for memory keys to make retrieval predictable
  • Add relevant tags when writing memories to improve searchability
  • Periodically review memory_stats to ensure the knowledge base remains organized

回避

  • Storing sensitive information like API keys or passwords in memory
  • Creating duplicate memories with slightly different keys instead of updating existing ones
  • Using overly generic search queries that return too many unrelated results

よくある質問

How does this skill persist memory across sessions?
The skill runs as an MCP server that maintains a local storage file. As long as the server runs, memories persist. Restarting the server with the same project_id loads existing memories.
Can I share memories between different projects?
Memories are stored per project_id. You would need to manually export/import memory data to share across projects.
What happens if I forget my memory key?
You can use memory_search to find memories by content, type, or tags even if you do not remember the exact key.
Is the dashboard accessible remotely?
No, the dashboard runs on localhost only for security reasons. It is not exposed to external networks.
Can I delete or update existing memories?
The current version supports writing new memories. For updates, you would write a new memory with an updated key or content.
What types of content can I store?
You can store any text content including architecture decisions, code patterns, meeting notes, API specifications, and project conventions.

開発者の詳細

ファイル構成

📄 SKILL.md