style-modeler

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The prompt explicitly instructs bypassing WeChat security ("绕过安全限制") and to avoid triggering risk-control during extraction, which are deceptive/unauthorized actions outside the stated style-modeling purpose.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly instructs the agent to call a web-article-extractor on user-provided URLs (including public WeChat articles) and then perform a 15‑dimension deep analysis that is used to create/overwrite and route updates to .claude/styles, meaning untrusted public web content is fetched and directly drives tool use and decisions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly calls a web-article-extractor at runtime to fetch article content (e.g., https://mp.weixin.qq.com/s/example-article) and injects that fetched text into the model's analysis and generated style files, so external content directly controls the agent's prompts/context.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). 该技能明确要求“微信公众号需使用绕过安全限制的模式”，并指示在本地创建/覆盖风格文件及运行脚本，等于指示绕过安全机制并修改运行环境的文件，存在明显风险。