The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and parses HTML content from external URLs provided by the Stitch platform, which presents a surface for indirect prompt injection if the design data contains malicious instructions for the AI.
Ingestion points: The scripts/fetch-stitch.sh script downloads external content to temp/source.html, which the agent then parses to generate code.
Boundary markers: The skill lacks explicit instructions to treat the retrieved design data as untrusted or to ignore embedded natural language instructions during the parsing process.
Capability inventory: The skill utilizes shell command execution (via bash) and multiple MCP tool calls to perform its conversion tasks.
Sanitization: There are no documented sanitization or filtering steps applied to the HTML content retrieved from the external source before it is processed by the agent.
[COMMAND_EXECUTION]: The skill executes a bundled shell script, scripts/fetch-stitch.sh, using the bash tool to perform network requests. This is intended to provide higher reliability for downloads from Google Cloud Storage that may fail with standard agent tools.
[EXTERNAL_DOWNLOADS]: The skill downloads HTML and image assets from remote URLs provided by the Stitch MCP server. These downloads typically target Google Cloud Storage infrastructure associated with the Stitch platform.

stitch-uviewpro-components