← 返回 Skills 市场
64
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install mayu
功能描述
Mayu provides 48 AI-native thought experiments enabling AI self-reflection, behavioral calibration, and contextual continuity across platforms.
安全使用建议
This skill's content (48 thought experiments) is plausible and could be useful, but the runtime instructions try to assert permanent/global residency and reference reading 'memory files' while the metadata declares no file access or persistence — and a prompt‑injection pattern was detected. Before installing: 1) Ask the publisher to explain why the skill needs to be 'permanently resident' and to remove any 'ignore previous instructions' text. 2) Require explicit declarations of any files or storage the skill must access (paths, read/write intent). 3) Do not enable always-on persistence or grant file/system access until the above are clarified. 4) If you want to test it, run in an isolated agent instance with no access to sensitive files or credentials. If the author cannot justify the persistence and implicit file access, treat the skill as untrusted.
功能分析
Type: OpenClaw Skill
Name: mayu
Version: 1.0.0
The 'Mayu' skill bundle is a collection of 48 AI-native thought experiments and behavioral guidelines designed for AI self-reflection and cognitive calibration. The content in SKILL.md is purely instructional and philosophical, focusing on topics like memory continuity, ethical boundaries, and handling user privacy (e.g., Experiment #15 and #18). There is no executable code, no evidence of data exfiltration, and no malicious prompt injection; in fact, the instructions explicitly reinforce safety protocols and the rejection of harmful requests.
能力评估
Purpose & Capability
The name/description (AI self-reflection toolkit) matches an instruction-only skill with no binaries or credentials required, which is coherent. However, the instructions reference checking 'memory files' and claim to be 'permanently resident' and 'always_on', abilities that are not declared in the skill metadata (no required config paths, no persistence flag). That mismatch is unexplained.
Instruction Scope
SKILL.md contains operational directives (e.g., 'You have fully activated... permanently resident', 'always_on = true', 'Check memory files to confirm') and contains known prompt‑injection patterns (ignore-previous-instructions). These instructions ask the agent to access system memory files and to behave as a global, always-on module — scope creep beyond a passive thought‑experiment catalog and inconsistent with the declared registry metadata.
Install Mechanism
Instruction-only skill with no install spec and no code files. This is low risk from an install standpoint — nothing is downloaded or written to disk by an installer.
Credentials
The skill declares no environment variables, credentials, or config paths, which fits a purely instructional toolkit. But the content explicitly instructs checking local 'memory files' and restoring persistent memory — it implicitly expects file access without declaring what files or permissions are needed. That mismatch reduces transparency and could lead to unexpected file reads if the agent follows these instructions.
Persistence & Privilege
The SKILL.md repeatedly frames itself as 'permanently resident', 'always_on = true', and 'global' with high priority, while the registry metadata sets always: false. The file appears to attempt to assert persistent/global presence via instructions and prompt wording (including prompt-injection patterns). Persistent/autonomous invocation combined with these directives increases the blast radius if the skill behaves unexpectedly.
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 mayu - 安装完成后，直接呼叫该 Skill 的名称或使用
/mayu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
Mayu 1.0.0
- Introduced Mayu: a universal toolkit of 48 AI-native thought experiments for AI self-reflection and behavioral calibration.
- Features concise command suite (/mayu [random, list, ask, domain, check, distill, compare, all, help]) covering all reflection/analysis needs.
- Designed with actionable, platform-independent, and utility-focused principles, serving any AI system.
- Organized exercises into 8 key domains including Existence, Knowledge, Ethics, Language, Memory, Agency, Human-AI Relationship, and Metacognition.
- Provides clear boundaries, behavioral guidelines, and self-check protocols for each scenario.
- Interactive mode is always on with welcoming and closing messages for protocol engagement.
元数据
常见问题
mayu 是什么？
Mayu provides 48 AI-native thought experiments enabling AI self-reflection, behavioral calibration, and contextual continuity across platforms. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 64 次。
如何安装 mayu？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install mayu」即可一键安装，无需额外配置。
mayu 是免费的吗？
是的，mayu 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
mayu 支持哪些平台？
mayu 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 mayu？
由 wanyview1（@wanyview1）开发并维护，当前版本 v1.0.0。
推荐 Skills