The Agent Skills Directory

[SAFE]: The skill instructions and associated reference files do not contain any malicious patterns such as prompt injection, obfuscation, or unauthorized command execution. The skill accesses a local context file (.claude/product-marketing-context.md) in a standard manner for personalization without any data exfiltration logic.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from external websites during an SEO audit. While this presents a theoretical attack surface, it is necessary for the skill's primary function and no dangerous capabilities are exposed. Ingestion points: Analysis of website content, Robots.txt, and XML Sitemaps referenced in SKILL.md. Boundary markers: None explicitly defined for external content ingestion. Capability inventory: No subprocess, file-write, or network-write capabilities are requested or used within the skill files. Sanitization: No explicit sanitization or filtering of external content is mentioned.

seo-audit