← 返回 Skills 市场

审核质检

Name: 审核质检
Author: jeyeshield

作者 JEyeshield · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

225

总下载

当前安装

版本数

在 OpenClaw 中安装

/install ad-production-review-quality

功能描述

审核质检技能 - 自动化质量评估和人工审核工作流

使用说明 (SKILL.md)

Review Quality - 审核质检

负责广告素材的质量评估和审核工作流。

Setup

无需额外依赖，TypeScript编译后使用。

When to Use

自动审核广告素材质量
人工审核流程管理
质量评估和打分
合规性检查

Architecture

review-quality/
├── index.ts          # 主入口，审核质检逻辑
├── package.json      # 依赖配置
└── README.md         # 详细文档

Core Commands

自动审核

await api.executeAction('review-quality.auto-review', {
  materialId: string,    // 素材ID
  criteria?: {          // 审核标准
    resolution?: { minWidth: number, minHeight: number },
    format?: string[],
    maxFileSize?: number
  }
});

人工审核

await api.executeAction('review-quality.manual-review', {
  materialId: string,    // 素材ID
  reviewer: string,      // 审核人
  decision: 'approved' | 'rejected' | 'revise',
  comments?: string
});

质量评估

await api.executeAction('review-quality.assess', {
  materialId: string,    // 素材ID
  dimensions: string[]   // 评估维度
});

获取审核状态

await api.executeAction('review-quality.get-status', {
  materialId: string     // 素材ID
});

响应事件

review-quality.auto-reviewed - 自动审核完成
review-quality.manual-reviewed - 人工审核完成
review-quality.assessed - 质量评估完成

安全使用建议

This skill's implementation appears harmless (no network calls or secret access) but the SKILL.md documentation and the actual code disagree about action names and event names. Before installing or using it: 1) Verify which API your agent expects (executeAction('review-quality.*') vs. registerCommand('review' subcommands) and update either the code or docs so they match. 2) Confirm the event names your system listens for — the code emits 'review.completed'/'review.rejected' while the documentation lists different event names. 3) Note that reviews are stored only in memory (Map) — they will be lost on restart; if persistent storage is needed, request that change. 4) Test the skill in a development environment and inspect emitted events to ensure they won't be accidentally handled by unrelated listeners. These inconsistencies look like sloppy documentation or bundling rather than malicious behavior, but they can cause silent failures, so fix/verify before production use.

功能分析

Type: OpenClaw Skill Name: ad-production-review-quality Version: 1.0.0 The skill bundle implements a standard quality review and assessment workflow for advertising materials. The code in index.ts uses in-memory storage for review records and provides typical CRUD and batch operations without any evidence of data exfiltration, unauthorized execution, or malicious prompt injection.

能力评估

ℹ Purpose & Capability

The implemented index.ts provides in-memory review storage, automated checks, manual review commands, batch operations, stats, and listens for generation.completed — which aligns with an ad review/quality workflow. However, SKILL.md and comments reference different action names and event names (e.g., SKILL.md shows api.executeAction('review-quality.auto-review') and events like 'review-quality.auto-reviewed'), whereas the code registers a top-level command named 'review' with subcommands (check/submit/...) and emits 'review.completed' / 'review.rejected'. This mismatch is an incoherence between docs and implementation.

ℹ Instruction Scope

SKILL.md instructs use of executeAction('review-quality.*') and documents certain event names, but the runtime code exposes command-based API and different event names. Aside from the naming mismatch, the instructions and code do not attempt to read files, environment variables, or contact external endpoints — no suspicious data-collection or exfiltration behavior is present.

✓ Install Mechanism

No install spec is provided (instruction-only skill); package.json lists only devDependencies (TypeScript, @types/node). There are no downloads, external installers, or extracted archives — low install risk.

✓ Credentials

The skill declares no required environment variables or credentials. The code briefly references api.user?.id (platform-provided user identity) but does not request or read secrets, tokens, or config paths — access appears proportionate for its function.

✓ Persistence & Privilege

always is false and disable-model-invocation is false (normal). The skill keeps data in an in-memory Map (no persistent storage), registers its own commands/events, and listens for generation.completed. It does not modify other skills or system-wide settings.

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 ad-production-review-quality
安装完成后，直接呼叫该 Skill 的名称或使用 /ad-production-review-quality 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

初始版本发布

元数据

Slug ad-production-review-quality

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

审核质检是什么？

审核质检技能 - 自动化质量评估和人工审核工作流. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 225 次。

如何安装审核质检？

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

审核质检是免费的吗？

是的，审核质检完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。

审核质检支持哪些平台？

审核质检跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。

谁开发了审核质检？

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