The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes technical project assets that could potentially contain hidden instructions designed to influence the agent's behavior.
Ingestion points: Downloads and parses external HTML/CSS source code and project metadata from the Stitch platform via the Stitch MCP server.
Boundary markers: The instructions do not include specific delimiters or instructions to ignore embedded prompts within the technical assets being analyzed.
Capability inventory: The skill utilizes the web_fetch tool to retrieve remote assets and the Write tool to generate the synthesized documentation file.
Sanitization: There is no mention of sanitizing or filtering the HTML source code before it is interpreted by the agent.
[EXTERNAL_DOWNLOADS]: The skill retrieves code and documentation from external platforms to perform its core functions.
Skill Installation: The installation instructions involve fetching a skill package from the google-labs-code/stitch-skills repository.
Asset Fetching: The workflow involves using web_fetch to download HTML source code and visual screenshots from dynamic URLs provided by the Stitch MCP server tools.
External Guidelines: References official design and prompting documentation hosted on the stitch.withgoogle.com domain.

design-md