The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user design requests and materials which are then interpolated into prompts for UI generation tools.
Ingestion points: User design intent and materials provided in the initial request, as described in SKILL.md and references/workflow.md.
Boundary markers: The skill uses structured section headers ([Context], [Layout], [Components]) to delimit generated prompt content.
Capability inventory: The skill orchestrates multiple MCP tools including generate_screen_from_text, edit_screens, generate_variants, and stitch-upload.
Sanitization: Clear instructions are provided to replace sensitive customer data (names, phone numbers, order IDs) with demonstration values and to avoid including credentials or session data in outputs.- [EXTERNAL_DOWNLOADS]: The skill references a specific snapshot of an official GitHub repository from Google (google-labs-code) as a source for its implementation logic and standard. This reference targets a well-known service provider and is handled neutrally.- [COMMAND_EXECUTION]: The skill is designed to orchestrate complex UI design workflows using specialized MCP tools. It implements defensive patterns such as idempotency checks (verifying remote state through get_project and list_screens before retrying failed writes) and visual verification of results via screenshot retrieval tools (get_screen).

stitch-ui-execute