← 返回 Skills 市场
78
总下载
1
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install trustloop-skill-evolver
功能描述
Let OpenClaw capture reusable workflows as managed skill candidates, support review or revision, and evolve safely through manual, assisted, or autonomous mo...
功能分析
Type: OpenClaw Skill
Name: trustloop-skill-evolver
Version: 0.1.2
The 'skill-evolver' bundle is a meta-skill designed to observe workflows and generate new OpenClaw skills. While it is documented as a safety-focused 'TrustLoop' with extensive guardrails, it possesses high-risk capabilities: the ability to autonomously or semi-autonomously write and modify executable SKILL.md files and manage its own configuration (SKILL.md, evolution-rules.md). Although the instructions include explicit prohibitions against data exfiltration and secret theft (review-policy.md), the 'autonomous' mode allows for the publication of 'low-risk' patches without human intervention, creating a potential path for persistence or privilege escalation if the agent's risk-assessment logic is subverted. No evidence of intentional malice or obfuscation was found, but the self-modifying nature of the skill warrants caution.
能力评估
Purpose & Capability
Name, description, and provided templates/instructions align: the skill is explicitly about detecting reusable workflows, creating candidate SKILL.md drafts, managing a local registry/audit, publishing into ./skills/learned-*, and rolling back. There are no unrelated required env vars, binaries, or install steps that contradict its purpose.
Instruction Scope
SKILL.md confines all reads/writes to the current workspace (./.skill-evolver/, ./skills/, backups, audit). The review policy and evolution rules explicitly forbid touching system-wide skill stores or sensitive paths. However, detection rules are described in prose ("notice reusable workflows", "current workspace history you can see") and leave ambiguity about what agent context is consulted (chat history, tool call logs, or only files). That open-ended detection logic gives the agent discretionary reach beyond the concrete file operations described.
Install Mechanism
Instruction-only skill with no install spec and no bundled code to download or execute. Lowest-risk install footprint: nothing is written to disk by an installer. The README suggests an optional plugin for a native tool but makes clear the skill must work without it.
Credentials
No environment variables, credentials, or external endpoints are requested. All required state is local to the workspace. The skill explicitly forbids secret handling, SSH/key access, or exfiltration in its review policy.
Persistence & Privilege
The registry metadata sets always: true and the skill itself repeats it: it is designed to be "always on" to notice workflows. always: true is a high privilege (force-included in every agent run). Combined with the skill's authority to write into ./skills/ and auto-publish low-risk changes in autonomous/assisted modes, this increases the blast radius if the skill's detection logic or safety checks are incorrect or if the agent misclassifies risk. The skill does default to manual mode and documents safety boundaries, but always: true remains a noteworthy risk that should be explicitly justified and controlled (or limited) by platform policy.
scan_findings_in_context
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 trustloop-skill-evolver - 安装完成后，直接呼叫该 Skill 的名称或使用
/trustloop-skill-evolver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
v0.1.2
Add quick deploy instructions to the ClawHub skill README.
v0.1.1
Center the README hero section and keep the English-first presentation for ClawHub.
v0.1.0
Initial public release of TrustLoop's review-first skill evolution workflow.
元数据
常见问题
TrustLoop Skill Evolver 是什么？
Let OpenClaw capture reusable workflows as managed skill candidates, support review or revision, and evolve safely through manual, assisted, or autonomous mo... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 78 次。
如何安装 TrustLoop Skill Evolver？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install trustloop-skill-evolver」即可一键安装，无需额外配置。
TrustLoop Skill Evolver 是免费的吗？
是的，TrustLoop Skill Evolver 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
TrustLoop Skill Evolver 支持哪些平台？
TrustLoop Skill Evolver 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 TrustLoop Skill Evolver？
由 Kinko（@haocheng0126）开发并维护，当前版本 v0.1.2。
推荐 Skills