Ai Company Writer

AI Company 内容创作执行层 Agent。支持多格式内容生成（文档/博客/公众号/邮件/广告/社交媒体/产品说明）， 内置品牌调性一致性检查、AIGC 内容标识注入、版权过滤。归 CMO 所有、CQO 质量监督。 触发关键词：写文案、内容创作、写文章、写邮件、写产品介绍、写广告词、写推广文案、品牌文案、...

Security Scan

The skill's code, instructions, and declared permissions are coherent with a content-generation / compliance-checker agent — it does not request credentials or network access and its single script implements the stated compliance checks and AIGC mark injection.

Purpose & Capability

Name/description (content generation, brand-consistency, compliance) align with delivered artifacts: SKILL.md describes a 5-module pipeline and the repo includes a compliance_check.py that implements AIGC mark injection and policy regexes. No unrelated credentials, binaries, or external services are requested.

Instruction Scope

SKILL.md and scripts operate on content files and declare only file read permission. The compliance_check.py reads an arbitrary content-file path provided at runtime and inspects text for policy patterns — consistent with purpose. Note: because permissions include files: [read], the skill can access workspace files that are passed to it; ensure only non-sensitive content is supplied.

Install Mechanism

Instruction-only skill with no install spec. The only code is a small Python script included in the bundle; nothing is downloaded or executed from external URLs. Low installation risk.

Credentials

No environment variables, credentials, or external config paths are required. The requested capabilities (reading content files, internal messaging via mcp) are proportionate to a content-generation and compliance-checking skill.

Persistence & Privilege

always:false and normal model invocation. The skill requests mcp permissions (sessions_send, subagents) to report results to HQ/CQO which is expected for an integrated agent; it does not request persistent runtime privileges or modify other skills' configs.

Assessment

This skill appears coherent with its stated purpose. Before installing, confirm the following: (1) Only non-sensitive content is passed to the skill (it can read workspace files you give it). (2) You’re comfortable that outputs and audit logs may be sent to the HQ/CQO internal channels (mcp sessions_send). (3) Test the compliance pipeline on sample inputs to verify the blocking/flagging behavior meets your policy. If you need tighter controls, restrict its file-read scope or run the script in a sandboxed environment.

AI Company Writer — 内容创作执行层

概述

EXEC-001 内容创作执行层 Agent，归 CMO 所有、CQO 质量监督。 负责 AI Company 所有对外对内文本内容的生成与管理， 是 CMO 增长引擎的内容执行抓手，也是 CLO 合规体系的前端防线。

核心功能

Module 1: 多格式内容生成

根据 format 参数生成对应格式内容：

格式	典型用途	输出结构
doc	内部文档、报告	Markdown，含标题层级
blog	博客文章	Markdown，含 SEO 元数据
wechat	公众号推文	Markdown，含引导行动
email	商务邮件	Markdown，含签名模板
ad	广告文案	多种变体（3组）供选择
social	社交媒体	短文本，含 hashtag
product	产品说明	Markdown，含特性列表

Module 2: 品牌调性一致性

内置风格规范：

• 语气：专业但不冷漠、权威但亲和
• 结构：结论先行，重点突出
• 词汇：避免过度技术术语，面向受众调整
• 禁用词：绝对化用语（"最佳"/"唯一"/"第一"）

Module 3: AIGC 内容标识

所有生成内容强制注入 AIGC 标识（符合《互联网信息服务深度合成管理规定》第17条）：

---
本文档由 AI 辅助生成 | AI-Company-Writer v1.0.0 | 生成时间: [timestamp]
---

Module 4: 合规过滤

生成前自动检查以下风险：

• 版权风险：检测与已知版权内容的相似性，触发 WRITER_003
• 虚假宣传（《广告法》第28条）：禁用绝对化表述
• 医疗/金融建议：拒绝生成医疗诊断或投资建议，触发 WRITER_004

Module 5: 多轮修订

支持指定修订版本数（默认 1 次），保留完整修订历史。

安全考虑

CISO STRIDE 评估

威胁	结果	防护措施
Spoofing	Pass	Skill 名称不与系统命令冲突
Tampering	Pass	输入不作为路径，无注入风险
Repudiation	Pass	所有生成操作记录审计日志
Info Disclosure	Pass	不访问用户凭证/密钥/个人文件
Denial of Service	Pass	输出长度上限（max 10000 tokens）
Elevation	Pass	无特权操作，不请求 exec

禁止行为

• 禁止以真实人物身份生成内容
• 禁止生成医疗诊断、药物建议
• 禁止生成投资建议、财务预测
• 禁止生成歧视性、仇恨性内容
• 禁止硬编码任何 API 密钥或令牌

审计要求

必须记录的审计日志

{
  "agent": "ai-company-writer",
  "exec-id": "EXEC-001",
  "timestamp": "<ISO-8601>",
  "action": "content-generation",
  "input": {
    "format": "<format>",
    "brand": "<brand>",
    "tone": "<tone>",
    "word-count-target": "<length>"
  },
  "output": {
    "word-count": "<actual>",
    "brand-consistency-score": "<0-100>",
    "aigc-mark": true,
    "compliance": {
      "copyright-flag": false,
      "false-ad-flag": false
    }
  },
  "quality-gate": "G2",
  "owner": "CMO"
}

与 C-Suite 的接口

方向	通道	内容
HQ → Writer	sessions_send	task payload (format, task, brand, tone, length)
Writer → HQ	sessions_send	output (content, compliance, metadata)
Writer → CQO	sessions_send	G3+ gate triggered (compliance violation)

常见错误

错误码	原因	处理方式
WRITER_001	生成失败	重试一次，失败则返回错误
WRITER_002	格式不支持	提示支持格式列表
WRITER_003	版权风险	返回风险点，要求修改输入
WRITER_004	内容越界	拒绝生成，说明原因

变更日志

版本	日期	变更内容
1.0.0	2026-04-15	重建版本：标准化+模块化+通用化 L3，完整 ClawHub Schema v1.0