OpenViking记忆适配器
v1.0.0字节OpenViking记忆系统适配器 - 分层加载优化,Token降低83%
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
Name/description (OpenViking memory adapter) match what the package does: reading agent memory files from ~/.openclaw/workspace, extracting key info, producing L0/L1/L2 artifacts, and exposing tools for analysis and search. The declared artifacts and tool names align with the code and skill.json.
Instruction Scope
SKILL.md describes local tool calls (openclaw tools call ...) and the server.py implements reading/writing files under ~/.openclaw/workspace and producing summaries. The instructions do not ask for unrelated system files or extra credentials. The code does read potentially sensitive local memory files (MEMORY.md, memory/*.md, USER.md, SOUL.md) — that is expected for a memory adapter but is also precisely the sensitive data surface the user should be aware of.
Install Mechanism
No install spec is provided (instruction-only from registry perspective) and the code is executed via MCP (python3 server.py). requirements.txt lists requests and python-dotenv but there is no external installer or downloaded binaries. This is low-install risk; however you should ensure dependencies are installed from trusted sources.
Credentials
The skill declares no required env vars, no primary credential, and no config paths beyond ~/.openclaw/workspace which it uses for reading/writing memory — these are proportional to a memory adapter. requirements.txt includes python-dotenv and requests (possible network usage) though the visible code does not access external env variables; that mismatch is minor but worth verifying.
Persistence & Privilege
Skill is not always:true and does not request elevated system privileges. It writes its own files under ~/.openclaw/workspace/memory_viking (expected). Autonomous invocation is allowed (platform default) but not combined here with other high-risk factors.
Assessment
This skill looks coherent for a local memory adapter, but before installing: (1) review the complete server.py for any network calls, remote endpoints, or code that sends memory content off-machine (the provided file was truncated); (2) confirm whether requests/dotenv are actually used and why; (3) remember it will read your agent's memory files (MEMORY.md, USER.md, SOUL.md, memory/*.md) and will write into ~/.openclaw/workspace/memory_viking — back up any sensitive data first; (4) if you run it, prefer executing in a sandboxed environment first to confirm no unexpected outbound traffic; (5) verify how the marketplace handles the stated per-call payment (0.5 USDT) and whether any credentials/payment hooks are required.Like a lobster shell, security has layers — review code before you run it.
latest
OpenViking Adapter
字节跳动开源 AI 记忆系统的 OpenClaw 适配实现。
核心特性
- 三层记忆架构: L0灵魂层 + L1概述层 + L2详情层
- 智能加载: 按需检索,避免信息过载
- Token优化: 降低83%的Token消耗
- 效果提升: 任务完成率35%→52%
工具功能
analyze_token_usage- 分析当前记忆系统的Token使用情况generate_l0_soul- 生成100token核心身份摘要generate_l1_overview- 生成2000token重要记忆概述search_relevant_memories- 智能搜索相关记忆optimize_memory_loading- 执行完整的优化流程
使用方法
- 安装:
clawhub install openviking-adapter - 分析:
openclaw tools call openviking-adapter analyze_token_usage - 优化:
openclaw tools call openviking-adapter optimize_memory_loading
定价
- 每次调用: 0.5 USDT
开源协议
基于字节跳动 OpenViking 开源项目
Comments
Loading comments...
