code-walkthrough

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill requires including raw filesystem command output (via showboat exec) into the generated document, which will verbatim capture any secrets present in files or command outputs (e.g., API keys, tokens, .env/credentials), creating a direct exfiltration risk.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill instructs installing and using Showboat (https://github.com/simonw/showboat and the git-style reference github.com/simonw/showboat@latest) at runtime, and Showboat runs arbitrary shell commands via showboat exec, so fetching and running that external repo would enable execution of remote code required by the skill.