The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content from the Mobbin website, which creates a surface for potential indirect prompt injection instructions to be ingested into the agent context.\n- Ingestion points: Step 2 involves retrieving search results and screenshots from mobbin.com.\n- Boundary markers: The skill does not define specific delimiters or instructions to ignore commands within the external data.\n- Capability inventory: The skill performs UI implementation and code generation in Step 6, which is a write-capable operation.\n- Sanitization: No specific sanitization or filtering of external visual or textual data is defined.\n- Mitigation: A mandatory human-in-the-loop approval step (Step 5) is required before any code implementation occurs, providing a critical safeguard against autonomous malicious actions.

mobbin-ux