The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted user-provided design briefs, external URLs, and screenshots to drive code generation, which creates a potential surface for indirect prompt injection.
Ingestion points: The agent is instructed to analyze signals from user briefs, URLs, and screenshots (SKILL.md, Section 0.A).
Boundary markers: The skill does not define specific delimiters or instructions for the agent to ignore potentially malicious embedded instructions within the user-provided data.
Capability inventory: The skill can generate React code, recommend package installations (npm/npx), and reference external scripts.
Sanitization: There are no specific instructions for sanitizing or filtering content extracted from the ingested briefs or external URLs.
[EXTERNAL_DOWNLOADS]: The skill references and fetches assets from external content delivery networks and repositories.
Evidence: The instructions reference Shopify's CDN for Polaris components and Simple Icons for SVG marks.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to generate or recommend the execution of shell commands for project initialization and package management.
Evidence: Includes instructions for using npx shadcn@latest init and various npm install commands for design system libraries.
[PROMPT_INJECTION]: The skill contains highly restrictive and specific instructions intended to override default model behavior and ensure a specific aesthetic output.
Evidence: Section 9.G explicitly mandates a "COMPLETELY banned" status for em-dash characters, describing it as a "non-negotiable" rule intended to override the model's historical behavior.

design-taste-frontend