The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest external data which creates a potential surface for indirect prompt injection attacks.
Ingestion points: The skill explicitly directs the agent to read .claude/product-marketing-context.md at the beginning of the task to gather context. It also processes user-supplied information about page content to generate schema markup.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat the content of the marketing context file as untrusted data or to ignore instructions embedded within it.
Capability inventory: The skill primarily outputs JSON-LD text blocks. It does not utilize system-level capabilities like file writing, network operations, or subprocess execution.
Sanitization: No specific sanitization or validation logic is defined within the skill instructions for the external data being processed, although it recommends using external validation tools (like Google Rich Results Test) for the final output.

schema-markup