The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to autonomously read task instructions from .stitch/next-prompt.md and visually styled guidelines from .stitch/DESIGN.md and interpolate them directly into tool prompts.
Ingestion points: Reads .stitch/next-prompt.md, .stitch/DESIGN.md, and .stitch/SITE.md using the Read tool.
Boundary markers: None identified; the content is concatenated into the prompt parameter for generate_screen_from_text.
Capability inventory: The skill has access to Bash, Write, and stitch*:* (generation/modification) tools.
Sanitization: No evidence of input validation or escaping for the ingested markdown content before it is sent to the Stitch MCP server.
[COMMAND_EXECUTION]: The skill uses Bash to start a local development server (e.g., npx serve site/public) for visual verification. While intended for testing, this provides an execution surface if a malicious iteration were to modify the site's scripts or dependencies.
[EXTERNAL_DOWNLOADS]: The skill downloads HTML and image assets from Stitch's infrastructure. These are expected within the context of the skill's purpose as a frontend builder and target the vendor's own resources.

stitch-loop