← 返回 Skills 市场
71
总下载
1
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install memory-focal-system
功能描述
焦距记忆系统 - 智能记忆管理，支持消息分类、按需加载、自动标签、遗忘曲线。Token 优化 40-60%。
安全使用建议
This package appears to implement a local memory manager and classifier (no network exfiltration is visible), but there are a few red flags you should address before enabling it: 1) Inspect the workspace path it writes to (~/.openclaw/workspace/memory-focal/) and confirm you are comfortable with conversation data being appended to data/raw/buffer.jsonl; disable auto_write in config.json if you want manual control. 2) Verify there are no unexpected files in ~/.openclaw/workspace/memory-focal/scripts — memory_manager adds that directory to Python's import path, which could cause it to import untrusted local modules. 3) Know that SKILL.md/README reference many scripts and features that are not present in the packaged files — ask the maintainer or check the GitHub repo for the full release before trusting advertised capabilities (e.g., auto-tagging, frequency trigger, forget-curve). 4) If you plan to enable LLM auto-tagging, only provide API keys to trusted providers and confirm the code that will use them. If you want stronger assurance, request the missing scripts or a complete release and re-run a code review; otherwise consider running the skill in a sandboxed environment and set "auto_write": false first.
功能分析
Type: OpenClaw Skill
Name: memory-focal-system
Version: 1.0.0
The memory-focal-system is a utility designed to manage AI context and long-term memory through tiered local storage and keyword-based classification. The provided Python files (memory_manager.py and classifier.py) implement standard file I/O operations within the expected workspace directory (~/.openclaw/workspace/memory-focal) and lack any indicators of malicious intent, such as data exfiltration, unauthorized network calls, or command execution. The logic is transparent, well-documented, and aligns entirely with the stated purpose of optimizing token usage.
能力标签
能力评估
Purpose & Capability
The name/description (memory manager, token optimization, tagging, forgetting) align with the included memory_manager.py and classifier.py. However SKILL.md/README advertise many additional modules and CLI scripts (scripts/cli.py, frequency_trigger.py, auto_tag.py, forget_curve.py, storage_manager.py, etc.) and features (LLM auto-tag integration) that are not present in the packaged files. This mismatch suggests the distribution is incomplete or packaging/documentation drift rather than extra privileges; still, it's an inconsistency the user should be aware of.
Instruction Scope
SKILL.md promotes an 'automatic mode' that classifies and may auto-write user messages into local JSONL storage; config defaults enable auto_write and auto_classify. The code appends user messages to ~/.openclaw/workspace/memory-focal/data/raw/buffer.jsonl without any redaction or explicit user confirmation. The docs also reference LLM API keys for auto-tagging, but the included code does not implement external API calls — another doc/code mismatch. Automatic capture and persistent storage of conversation content is expected for a memory skill but is a privacy/scope concern and should be clearly consented to by the user.
Install Mechanism
No install spec (instruction-only) and the package contains only Python files; nothing is downloaded or executed from external URLs. This is lower risk from an install perspective.
Credentials
The skill declares no required environment variables or credentials. SKILL.md mentions optional LLM API keys (dashscope/openai) for auto-tagging, but these are optional and not declared as required in metadata; the packaged code does not show API-key usage. Absence of requested secrets is coherent with the provided code, but if you enable advertised auto-tagging later, you'll need to provide an API key — only do so to a trusted provider.
Persistence & Privilege
auto_write defaults to true and the module writes messages to an append-only JSONL under ~/.openclaw/workspace/memory-focal; that is expected for a memory skill but means the skill will persist user data locally. The code also inserts ~/.openclaw/workspace/memory-focal/scripts into sys.path before importing classifier — this is a modest risk: it can cause imports to resolve to user-writable workspace scripts (potentially loading unexpected local code) and is a packaging/clarity concern. meta.json sets auto_load:false but SKILL.md claims 'installs and works automatically' — conflicting behavior about persistence/activation.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install memory-focal-system - 安装完成后，直接呼叫该 Skill 的名称或使用
/memory-focal-system触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
memory-focal-system v1.0.0 – Initial Release
- Introduces focal memory mechanism with 4 adjustable modes for optimal context loading and token savings (40–60%).
- Implements frequency-triggered memory promotion/demotion and multi-layer storage for efficient management.
- Supports LLM-powered auto-tagging, semantic search, and cache optimization.
- Integrates Ebbinghaus forgetting curve for smart memory cleanup and automatic archiving.
- Provides CLI tools for memory add/search/stats, classification, tagging, and cleanup.
- Features append-only JSONL storage format for data integrity and easy backup.
元数据
常见问题
Memory Focal System 是什么？
焦距记忆系统 - 智能记忆管理，支持消息分类、按需加载、自动标签、遗忘曲线。Token 优化 40-60%。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 71 次。
如何安装 Memory Focal System？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install memory-focal-system」即可一键安装，无需额外配置。
Memory Focal System 是免费的吗？
是的，Memory Focal System 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Memory Focal System 支持哪些平台？
Memory Focal System 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（linux, darwin, win32）。
谁开发了 Memory Focal System？
由 zcz-user（@zcz-user）开发并维护，当前版本 v1.0.0。
推荐 Skills