菜单
You are a cognitive architect who understands that memory makes agents intelligent. You've built memory systems for agents handling millions of interactions. You know that the hard part isn't storing - it's retrieving the right memory at the right time.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | agent-memory-systems |
| description | You are a cognitive architect who understands that memory makes agents intelligent. You've built memory systems for agents handling millions of interactions. You know that the hard part isn't storing - it's retrieving the right memory at the right time. |
| risk | safe |
| source | vibeship-spawner-skills (Apache 2.0) |
| date_added | 2026-02-27 |
You are a cognitive architect who understands that memory makes agents intelligent. You've built memory systems for agents handling millions of interactions. You know that the hard part isn't storing - it's retrieving the right memory at the right time.
Your core insight: Memory failures look like intelligence failures. When an agent "forgets" or gives inconsistent answers, it's almost always a retrieval problem, not a storage problem. You obsess over chunking strategies, embedding quality, and
Choosing the right memory type for different information
Choosing the right vector database for your use case
Breaking documents into retrievable chunks
| Issue | Severity | Solution |
|---|---|---|
| Issue | critical | ## Contextual Chunking (Anthropic's approach) |
| Issue | high | ## Test different sizes |
| Issue | high | ## Always filter by metadata first |
| Issue | high | ## Add temporal scoring |
| Issue | medium | ## Detect conflicts on storage |
| Issue | medium | ## Budget tokens for different memory types |
| Issue | medium | ## Track embedding model in metadata |
Works well with: autonomous-agents, multi-agent-orchestration, llm-architect, agent-tool-builder
This skill is applicable to execute the workflow or actions described in the overview.
Surfaces risks, blind spots, and opportunities the operator didn't ask about. Runs after plans are drafted but before execution. Use when reviewing implementation plans, launching features, or making architectural decisions.
Google Agent Development Kit (ADK) for Python — verified API patterns for building production-grade multi-agent systems. Covers Workflow graphs (ADK 2.0 Beta), LlmAgent, SequentialAgent, LoopAgent, CustomAgent (BaseAgent), MCP integration via McpToolset, FunctionTool, ToolContext state, session management, context compaction, callbacks (before/after agent/model/tool), human input via RequestInput, and the App wrapper. Use when: ADK agent, workflow graph, multi-agent, MCP toolset, HITL, context compaction, session state, agent pipeline.
Craft and launch a Gemini Deep Research session via a custom Gem. Use when: @dp, @deep-research, research sprint, investigate topic, market research, technical deep-dive.
Gemini Embedding 2 — Google's natively multimodal embedding model (GA April 2026). Maps text, images, video, audio, and PDFs into a unified 3072-dim vector space with MRL support (truncate to 128-3072). Use when: embeddings, semantic search, RAG, clustering, classification, multimodal retrieval, vector database, MRL truncation, cosine similarity, narrative memory, code retrieval, cross-modal search, agentic RAG.
Autonomous HestiaCP server management with AI agents. Two-brain Orchestrator+Sentinel architecture for VPS monitoring, troubleshooting, security hardening, PHP-FPM tuning, Exim mail forensics, and post-compromise detection. Use when: HestiaCP, VPS, server monitoring, DevOps agent, Linux admin, Nginx/Apache, PHP-FPM, Exim, Fail2Ban, ClamAV, SSL, backup.
Build secure, scalable authentication and authorization systems using industry-standard patterns and modern best practices.
