The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs and executes shell commands including npx shadcn@latest add <component>, npx remotion render, and node -e. These commands use parameters like <component> and <flow-slug> derived from user input or agent-generated plans, which could lead to command injection if input is not properly sanitized.
[DATA_EXFILTRATION]: In the live-components mode, the skill recursively reads files from user-specified paths. This capability could result in the exposure of sensitive data (such as .env files or credentials) if the agent is instructed to read from non-component directories.
[PROMPT_INJECTION]:
Ingestion points: Reads existing source code files from the project directory in live-components mode.
Boundary markers: None identified for distinguishing code from instructions.
Capability inventory: File system write access, shell command execution (npx, node, vite), and dynamic code generation.
Sanitization: No validation or sanitization of file content is specified before processing.
Risk: The skill is vulnerable to indirect prompt injection where malicious code comments or strings in processed project files could influence agent behavior.
[EXTERNAL_DOWNLOADS]: The skill downloads and executes tools and libraries from the npm registry, including shadcn, remotion, daisyui, and vite. These are well-known services and standard tools in the web development ecosystem.

ui-flow-studio