The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves and analyzes HTML source code from external project URLs. This data is considered untrusted as it originates from the project content being analyzed, creating a surface for embedded instructions to influence the agent.
Ingestion points: SKILL.md (Retrieval and Networking section, step 5) instructs the agent to download HTML from htmlCode.downloadUrl and screenshots from screenshot.downloadUrl.
Boundary markers: The skill does not implement delimiters or specific instructions to ignore potential commands embedded within the retrieved project data.
Capability inventory: The skill is configured with web_fetch for network access and Write for file system output (generating the DESIGN.md file).
Sanitization: No sanitization or filtering logic is defined for the content extracted from external project assets.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch design tokens, project metadata, and documentation from Google-hosted domains (stitch.withgoogle.com). These operations are consistent with the skill's primary function and target well-known services.

design-md