The Agent Skills Directory

[SAFE]: The skill defines a standard workflow for iterative UI design within the Stitch ecosystem. All instructions are focused on guiding the model to provide high-quality prompts to specific internal tools.
[DATA_EXPOSURE]: No hardcoded credentials, API keys, or access to sensitive local file paths (e.g., .ssh, .aws) were detected.
[REMOTE_CODE_EXECUTION]: The skill does not attempt to download external scripts or execute untrusted code from remote sources. Although the 'Bash' tool is permitted in the frontmatter, the instructions do not use it for malicious purposes.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied arguments and screen data to compose prompts for the 'edit_screens' tool. This establishes a known surface for indirect prompt injection where data from a screen might influence the agent's behavior. However, this is inherent to the tool's primary function (content refinement), and the instructions include best practices and templates to structure these interactions safely. No sanitization logic is explicitly defined in the markdown, but the capability remains within the scope of the intended user workflow.

stitch-iterate