← 返回 Skills 市场
jiuge897

Memory Pruner

作者 jiuge897 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
271
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install memory-pruner
功能描述
Manages agent memory by auto-pruning old or low-relevance entries, compressing duplicates, and reporting storage costs with safe defaults.
安全使用建议
Do not install or run this skill without verifying where the 'memory-pruner' CLI comes from. The SKILL.md instructs file reads and deletions but the package does not include the referenced executable or memory/ directory; an agent would rely on a local or external binary of unknown provenance. Before using: (1) inspect and obtain the memory-pruner implementation from a trusted source, (2) run only dry-run/analyze first and inspect suggested deletions, (3) ensure backups are stored in a safe location you control, (4) restrict the agent’s permissions so it cannot delete unrelated files, and (5) prefer manual confirmation before any prune operation. If you cannot verify the CLI origin or prefer not to grant filesystem deletion rights, consider declining this skill.
功能分析
Type: OpenClaw Skill Name: memory-pruner Version: 1.0.0 The skill bundle describes a utility for managing and optimizing agent memory by pruning old or low-relevance entries. The documentation (SKILL.md) and configuration (config.json) are consistent with this stated purpose and contain no evidence of malicious intent, data exfiltration, or prompt injection. While the primary executable script 'memory-pruner' was not included in the provided files, the metadata and instructions appear legitimate and safe.
能力评估
Purpose & Capability
The name, description, and commands (prune, analyze, compress, dry-run) coherently describe a memory management tool. However, the SKILL.md and config.json reference a 'memory-pruner' CLI and a 'memory/' working directory as part of the skill's files, yet the package contains no executable or code files. That leaves the agent instructed to run a binary that is not provided by the skill bundle.
Instruction Scope
The runtime instructions explicitly instruct reading, analyzing, compressing, backing up (.bak), and deleting workspace memory files. Those operations are within the claimed purpose, but because this is an instruction-only skill with no implementation, an agent following these instructions could run any local 'memory-pruner' binary present on the system or an attacker-supplied binary. The SKILL.md promises 'never deletes without confirmation', but instructions still include destructive actions (prune) and file I/O — the absence of an included implementation and of strict path constraints raises risk.
Install Mechanism
No install spec is provided (lowest install risk). However, config.json lists 'entry':'memory-pruner' and 'runtime':'shell', implying an entrypoint that does not exist in the bundle. This mismatch means the skill expects an external CLI to be present; the origin and integrity of that CLI are undefined.
Credentials
The skill requests no environment variables, no credentials, and no config paths. That is proportionate to the stated purpose. There is no request for unrelated secrets or external service keys.
Persistence & Privilege
The skill is not always-enabled and uses normal autonomous invocation defaults. It does not request persistent system-wide configuration changes in its bundle. The main concern is not privilege flags but the potential for destructive filesystem operations if an external 'memory-pruner' binary is invoked.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install memory-pruner
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /memory-pruner 触发
  4. 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
Initial release of Memory Pruner—intelligent memory management for agents. - Auto-prunes old memories based on configurable thresholds. - Scores and retains more relevant memories, merging similar ones for efficiency. - Tracks memory usage costs and provides actionable analysis and recommendations. - CLI includes prune, analyze, compress, config, and dry-run commands. - Features safe deletion (confirmation and backup) and category-based organization.
元数据
Slug memory-pruner
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Memory Pruner 是什么?

Manages agent memory by auto-pruning old or low-relevance entries, compressing duplicates, and reporting storage costs with safe defaults. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 271 次。

如何安装 Memory Pruner?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install memory-pruner」即可一键安装,无需额外配置。

Memory Pruner 是免费的吗?

是的,Memory Pruner 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Memory Pruner 支持哪些平台?

Memory Pruner 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 Memory Pruner?

由 jiuge897(@jiuge897)开发并维护,当前版本 v1.0.0。

💬 留言讨论