The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user inputs such as 'briefs' and 'reference signals' (URLs, screenshots). These inputs are used to drive the design process and code generation. There are no explicit boundary markers or sanitization steps mentioned to prevent the agent from obeying malicious instructions embedded within these inputs.
Ingestion points: Section 0 (Brief Inference) in SKILL.md.
Boundary markers: Absent.
Capability inventory: Package installation (Section 2.A), file creation (Section 3.A), and image generation (Section 4.8) across SKILL.md and generated code.
Sanitization: Not specified.
[DYNAMIC_EXECUTION]: The skill provides several code skeletons and templates (Section 5.A, 5.B, 5.C) in SKILL.md that the agent is instructed to use and adapt at runtime. It also uses CLI tools like 'npx shadcn@latest' to dynamically add components to the project. This is a common pattern for developer-focused skills but represents an automated code generation surface.
[EXTERNAL_DOWNLOADS]: The skill references and downloads assets from several external, well-known services. This includes placeholder images from 'picsum.photos' and SVG icons from 'simpleicons.org'. It also provides links to numerous official documentation sites and GitHub repositories for established design systems (e.g., Google, Microsoft, IBM, Atlassian). These sources are recognized as trusted or well-known, and the downloads are documented neutrally as intended functionality.

design-taste-frontend