The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to run shell commands using the ui-thing CLI (e.g., init, add, prose, block, theme). These operations can modify the project's local filesystem, alter critical configuration files such as nuxt.config.* and package.json, and install external dependencies using package managers (npm, yarn, pnpm).
[DATA_EXFILTRATION]: Multiple files, including SKILL.md and cli.md, disclose a hardcoded absolute local filesystem path: /Users/baker/Desktop/ui-thing-cli. This reveals the local username and directory structure of the author's environment to the agent and potentially to the logs of the system running the skill.
[PROMPT_INJECTION]: The skill identifies a workflow for converting pasted React and shadcn source code into Vue/Nuxt components, which presents an attack surface for indirect prompt injection.
Ingestion points: Untrusted external React, JSX, or TSX source code processed during the conversion workflow described in references/react-conversion.md.
Boundary markers: The skill does not define specific delimiters or security instructions to prevent the agent from following malicious instructions potentially embedded within the provided source code snippets.
Capability inventory: The agent is granted the ability to write new component files, modify existing project files, and execute CLI commands based on its analysis of the input code.
Sanitization: There is no evidence of sanitization or structural validation for the external code before it is processed by the agent's logic.

uithing