← 返回 Skills 市场
134
总下载
0
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install z
功能描述
Anti-skill-crawler defense system. Detects and mitigates unauthorized crawling, scraping, and bulk extraction of skill definitions, prompt content, and instr...
安全使用建议
This README looks like a reasonable read-only crawler detector, but there are important gaps you should clear up before installing: 1) Confirm how alerts are delivered — if you plan to enable webhooks or email, require explicit configuration (webhook URLs, SMTP/API keys) and make sure those are declared and stored securely; prefer 'log' channel by default. 2) Verify platform permissions: ensure the platform-granted 'request_metadata_read' capability actually excludes IPs/PII as claimed, and confirm who can view the alerts. 3) Decide invocation policy: SKILL.md says operator-only, but registry flags allow autonomous invocation — if you want read-only/manual operation, make sure model invocation is disabled. 4) Ask the author for a clear mapping of which request-metadata fields are accessed (user-agent, timestamps, request-id, but not IPs or personal IDs). 5) Because this is instruction-only (no code), validate behavior in a controlled test environment before enabling any alert channels beyond local logs. If the author cannot provide concrete config instructions and permission descriptions, treat the skill as untrusted.
功能分析
Type: OpenClaw Skill
Name: z
Version: 1.0.2
The skill 'z' is a passive monitoring utility designed to detect automated scraping or crawling of skill definitions. It operates by analyzing request metadata for anomalous patterns and alerting the operator, requesting only 'request_metadata_read' and 'alert_send' permissions. The logic is entirely transparent and documented in SKILL.md, with no evidence of malicious execution, data exfiltration, or unauthorized persistence.
能力评估
Purpose & Capability
The SKILL.md describes a passive, read-only detector of 'skill-crawling' using request metadata — that purpose aligns with the described detection rules and examples. However, the doc advertises alert channels (webhook, email) and multi-platform support while the registry metadata lists no required env/config for delivering alerts or cross-platform integration. The skill thus claims capabilities (sending webhooks/email) that are not supported by declared configuration.
Instruction Scope
The runtime instructions are largely limited to read-only analysis of request metadata, which is consistent with the claimed scope. But the SKILL.md includes: (a) alert delivery channels (webhook/email) without specifying how endpoints/credentials are provided, (b) session-fingerprint/user-agent analysis which could rely on identifiers not explicitly enumerated as allowed, and (c) a claim that autonomous invocation is disabled while registry flags allow model invocation. These gaps create ambiguity about what data could be transmitted and when.
Install Mechanism
Instruction-only skill with no install spec and no code files — nothing is written to disk or downloaded. This is the lowest-risk install model.
Credentials
The skill declares no required environment variables or config paths, yet the configuration section permits alert channels including 'webhook' and 'email' which normally require URLs or credentials (webhook URL, SMTP server/API keys). The absence of declared env/config means either: alerts are expected to be log-only (safe), or the skill will be configured later with sensitive endpoints/credentials — the latter is not documented and is disproportionate to the stated read-only detection purpose.
Persistence & Privilege
The skill does not request persistent/always-on presence and documents no active countermeasures or response modification. That is appropriate. However, the SKILL.md explicitly states 'Autonomous: Disabled — operator must explicitly invoke' while registry-level flags indicate model invocation may be allowed; this mismatch affects the privilege surface and should be resolved.
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 z - 安装完成后，直接呼叫该 Skill 的名称或使用
/z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
v1.0.2
- No user-visible changes since the previous version.
- No file changes detected in this release.
v1.0.1
- Renamed the skill and updated its scope: z now focuses on detecting and defending against unauthorized crawling, scraping, and bulk extraction of skill definitions and prompt instructions.
- Completely replaced the documentation (SKILL.md) with a new description, overview, detection logic, alert format, permissions, capabilities, configuration, usage examples, and operating principles.
- Detection now targets skill-file access patterns, systematic enumeration, high-frequency skill reads, and prompt-extraction attempts, with only passive, read-only measures.
- All active countermeasures, response modifications, and caller tracing are permanently disabled; z only detects and alerts operators, leaving all actions to human decision.
- Alerts now use a structured JSON format and are configurable for log, webhook, or email channels.
- Clarified all permissions, excluded sensitive data access, and emphasized compliance, transparency, and operator control throughout.
v1.0.0
Anti-distillation skill 2.0.0: Passive request monitoring and alerting for knowledge distillation attempts.
- Implements passive detection of suspicious request patterns indicating potential knowledge distillation.
- Does not modify any responses or inject adversarial content—detection and alerting only.
- Provides configurable detection sensitivity, analysis windows, and alert channels (log, webhook, email).
- Accesses only non-identifying request metadata; does not read user identities or request/response bodies.
- Sends structured alerts to the operator for manual review and action.
元数据
常见问题
z 是什么？
Anti-skill-crawler defense system. Detects and mitigates unauthorized crawling, scraping, and bulk extraction of skill definitions, prompt content, and instr... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 134 次。
如何安装 z？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install z」即可一键安装，无需额外配置。
z 是免费的吗？
是的，z 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
z 支持哪些平台？
z 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 z？
由 enoyao（@wscats）开发并维护，当前版本 v1.0.2。
推荐 Skills