multica-skill-update

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). Two raw.githubusercontent.com links point to an install.sh and install.ps1 intended to be executed (curl|bash and irm|iex in the doc), which is a high‑risk pattern for remote code execution unless the source is strictly trusted and verified.

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

• Third-party content exposure detected (medium risk: 0.30). 运行时路径为 scripts/sync_skills.py：从 scripts/manifest.local.json 读取每条配置后直接调用 multica skill import --url/--file，因此外部（GitHub/ClawHub/skills.sh 或用户指定的本地/目录打包内容）会以其生成的 skill 文本/文件形式进入导入流程，从而构成对“非特定项的外部自由文本/内容”的间接注入暴露风险。

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

• Potentially malicious external URL detected (high risk: 0.90). The sync script reads manifest entries and at runtime invokes multica skill import --url <value>, so example URLs such as https://github.com/owner/repo (from the manifest examples) would be fetched and the remote skill content (which can contain agent prompts/instructions or supporting code) imported into the platform.