← 返回 Skills 市场
Unified Self Improving
作者
Cui Ruochen
· GitHub ↗
· v1.0.4
· MIT-0
240
总下载
0
收藏
0
当前安装
5
版本数
在 OpenClaw 中安装
/install unified-self-improving
功能描述
统一自我进化系统，整合 self-improving-agent、self-improving、mulch 三个技能的优势，提供结构化日志、三层存储、自动升级、模式检测、命名空间隔离和 token 高效的 JSONL 格式支持。
安全使用建议
This skill appears coherent and implements a local, file-based memory/learning manager. Before installing, review and optionally run the scripts in a sandbox: 1) verify the repository origin (homepage is missing), 2) ensure you have required CLI tools (jq, grep, etc.), 3) note that data is stored under ~/.openclaw/workspace/memory by default (or wherever MEMORY_ROOT points) — back up any important data, and be cautious if you set MEMORY_ROOT to a different path, 4) the namespace delete command uses rm -rf with an interactive prompt — pay attention to the prompt, and 5) if you need network isolation, run the skill in an environment without network access. If you want higher assurance, inspect the listed scripts line-by-line (they are plain shell) or run them with --help and in a disposable directory first.
功能分析
Type: OpenClaw Skill
Name: unified-self-improving
Version: 1.0.4
The bundle implements a tiered memory management system (HOT/WARM/COLD) for AI agents, allowing them to log corrections, errors, and patterns to local JSONL and Markdown files. The system includes logic for session management, namespace isolation, and automatic data migration between storage tiers based on access frequency. All operations are confined to the local workspace directory (~/.openclaw/workspace/memory), and the code uses standard Bash utilities (jq, grep, mv) without any evidence of network exfiltration, unauthorized execution, or intentional backdoors. The 'self-improving' aspect is handled through structured data storage rather than risky code generation.
能力评估
Purpose & Capability
Name/description describe a local self-improvement/memory manager and the included scripts implement exactly that: CLI commands to log/query/move/index/namespace/manage sessions and detect patterns. There are no extraneous credentials, remote endpoints, or unrelated binaries required.
Instruction Scope
SKILL.md and the scripts direct the agent to operate on local files under ~/.openclaw/workspace/memory (index.jsonl, hot/warm/cold JSONL/MD files). The instructions and code only read/write those paths and do not instruct collection or transmission of unrelated system data or secrets.
Install Mechanism
There is no install spec that downloads arbitrary code at install time. The repository includes shell scripts and docs only; those scripts run locally and do not fetch remote artifacts. Risk is low, but the scripts assume standard CLI tools (jq, grep, mv, mktemp, etc.), so ensure required utilities are present and trusted.
Credentials
The skill declares no required environment variables or credentials. The code does honor optional environment overrides MEMORY_ROOT and SKILL_ROOT which can change where data is stored—this is functional but worth noting because an env override could redirect storage to another location if misset.
Persistence & Privilege
always: false and model invocation is allowed (platform default). The skill only writes to its own workspace and memory directories; it does not modify other skills or system-wide agent configs. Namespace deletion prompts for confirmation before rm -rf.
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 unified-self-improving - 安装完成后，直接呼叫该 Skill 的名称或使用
/unified-self-improving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
v1.0.4
unified-self-improving 1.0.4
- Migrated all CLI and workflow documentation to the new unified `unified-self-improving <command>` interface, deprecating legacy skill-specific commands for clarity and consistency.
- Updated storage model to consolidate three learning layers, clarifying directory structures for HOT, WARM, and COLD.
- Improved configuration documentation (in SKILL.md) to describe core tunable environment variables and simplified the configuration path.
- Added unified multi-purpose commands (log, move, recall, namespace, index, import, session, detect-pattern), merging previously separate abilities and triggers into concise subcommands.
- Streamlined workflow and session management instructions; clarified command usage examples for logging, querying, moving, and importing data.
v1.0.3
**No functional changes. Documentation only.**
- Updated README.md.
- No code or logic changes in this version.
v1.0.2
- Documentation updated in README.md; no code or logic changes.
- No modifications to behavior or configuration.
- Version remains functionally consistent with previous release.
v1.0.1
- Added initial implementation scripts covering command utilities and storage library for the unified self-improving system.
- Introduced shell scripts for key functions: pattern detection, importing, indexing, logging, moving data, namespace management, querying, recalling, and session handling.
- Included supporting libraries for configuration, indexing, namespace, and storage management.
- No existing files were modified; this release seeds the full foundational script set.
v1.0.0
Initial release of "unified-self-improving" skill, integrating three self-improvement tools into a single, unified system:
- Combines the core strengths of self-improving-agent, self-improving, and mulch, including structured logs, multi-layered storage, efficient JSONL support, and namespace isolation.
- Offers a three-tier storage model (HOT/WARM/COLD) with automated upgrade/demotion rules and dual-format (Markdown & JSONL) in the HOT tier.
- Defines unified triggers: session end, errors, corrections, pattern detection, manual trigger, and heartbeat.
- Maintains unique interfaces and command compatibility from the integrated tools for advanced operations (e.g., promote, pattern detection, namespace management).
- Optimizes for storage efficiency, fast recall, and resilient data organization using pattern-based indexing and namespace isolation.
元数据
常见问题
Unified Self Improving 是什么？
统一自我进化系统，整合 self-improving-agent、self-improving、mulch 三个技能的优势，提供结构化日志、三层存储、自动升级、模式检测、命名空间隔离和 token 高效的 JSONL 格式支持。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 240 次。
如何安装 Unified Self Improving？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install unified-self-improving」即可一键安装，无需额外配置。
Unified Self Improving 是免费的吗？
是的，Unified Self Improving 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Unified Self Improving 支持哪些平台？
Unified Self Improving 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Unified Self Improving？
由 Cui Ruochen（@421zuoduan）开发并维护，当前版本 v1.0.4。
推荐 Skills