The Agent Skills Directory

[PROMPT_INJECTION]: The skill defines an automated multi-step chain workflow for batch generation that is susceptible to Indirect Prompt Injection.
Ingestion points: According to SKILL.md, the agent is instructed to parse output_components from the generate_screen_from_text tool response to find continuation suggestions.
Boundary markers: There are no instructions to validate or sanitize these suggestions, nor are there delimiters to separate the tool's structural output from potentially malicious content passed through from the initial prompt.
Capability inventory: The agent uses the generate_screen_from_text tool to execute these instructions.
Sanitization: The skill explicitly instructs the agent to bypass user review: "Automatically call generate_screen_from_text again with the suggestion text as the prompt — the user already initiated generation, no need to re-confirm."
[DATA_EXPOSURE]: The example files (examples/desktop.md, examples/mobile.md) include a hardcoded projectId ("3780309359108792857"). While likely a dummy value for documentation purposes, it represents a static identifier being promoted for use across different user sessions.

stitch-mcp-generate-screen-from-text