The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill mandates the global installation of the '@superdesign/cli' package from an external registry using 'npm install -g @superdesign/cli@latest'.
[REMOTE_CODE_EXECUTION]: The skill fetches remote instruction files from 'https://raw.githubusercontent.com/superdesigndev/superdesign-skill/main/skills/superdesign/INIT.md' and 'https://raw.githubusercontent.com/superdesigndev/superdesign-skill/main/skills/superdesign/SUPERDESIGN.md' at runtime, instructing the agent to follow the instructions contained within these files. This allows for arbitrary behavior modification via remote content.
[COMMAND_EXECUTION]: The skill executes multiple shell commands including 'npm install', 'superdesign --version', 'superdesign login', and several complex design management commands. It also instructs the agent to automatically perform initialization tasks without user confirmation.
[DATA_EXFILTRATION]: The skill performs an extensive codebase analysis, reading full source code from project files and passing it to the external 'superdesign' CLI. It also includes a tool to fetch external brand guides from user-provided URLs.
[PROMPT_INJECTION]: The skill includes instructions to bypass user review, such as 'Do NOT ask the user to do this manually — just do it'.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via the analysis of repository code. 1. Ingestion points: Full repository source code (JSX, TSX, CSS, config files). 2. Boundary markers: None identified. 3. Capability inventory: Shell command execution, global package installation, and remote file fetching. 4. Sanitization: None identified for the ingested source code context.

superdesign