The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from project-specific context files (e.g., .agents/product-marketing-context.md) and external websites via the WebFetch tool. This creates a surface for indirect prompt injection attacks where malicious instructions could be embedded in the processed content.
Ingestion points: The skill explicitly instructs the agent to read .agents/product-marketing-context.md or .claude/product-marketing-context.md and uses the WebFetch tool for gathering site information.
Boundary markers: There are no defined delimiters or specific instructions provided to the agent to disregard potentially malicious commands contained within these external sources.
Capability inventory: The skill is scoped to the Read and WebFetch tools as defined in the frontmatter.
Sanitization: The instructions do not include methods for sanitizing, validating, or filtering the content retrieved from external files or URLs.

30x-seo-architecture