The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local bash script scripts/fetch-stitch.sh with a download URL as an argument. The URL is retrieved dynamically from the Stitch API (htmlCode.downloadUrl). If the external service provides a maliciously crafted URL containing shell metacharacters, it could lead to command injection during the execution phase.
[PROMPT_INJECTION]: The skill processes data from external sources (Stitch API and downloaded HTML) which constitutes an indirect prompt injection surface.
Ingestion points: Untrusted data enters the agent context through the Stitch API (via get_project and get_screen) and the downloaded file temp/source.html in SKILL.md.
Boundary markers: The instructions do not define delimiters or specific "ignore embedded instructions" warnings for the agent when parsing the external design content.
Capability inventory: The skill utilizes the Bash tool for script execution and the Write tool for creating CSS and Markdown artifacts across all processing steps.
Sanitization: The skill lacks explicit sanitization or validation logic for the content fetched from the remote Stitch platform before it is used to generate code.

stitch-design-system