The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized commands were detected within the skill files.
[DATA_EXPOSURE]: The skill instructions suggest reading local project context from .agents/product-marketing-context.md or .claude/product-marketing-context.md. This is a standard practice for project-aware agent skills to gather necessary business context and does not involve external data exfiltration.
[INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface by ingesting user-provided site descriptions and local project context files. However, since the skill does not possess capabilities for network operations, command execution, or file system modifications, the risk associated with this surface is negligible.

site-architecture