The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data such as product specifications and customer personas to generate marketing copy, which presents a surface for indirect prompt injection.\n- Ingestion points: Product specifications, target personas, and project files accessed via the read_file and search_code tools.\n- Boundary markers: Absent. The instructions do not define specific delimiters to isolate external content or warn the agent to ignore instructions within that content.\n- Capability inventory: The agent has permissions to use write_file, create_file, and edit_file, allowing it to modify the codebase based on processed inputs.\n- Sanitization: Partial. The skill includes security guardrails (ASI01 and ASI09) that instruct the agent to reject prompts attempting to inject false claims and to ensure all social proof is traceable.

write-copy