The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from multiple sources including live web URLs, YouTube videos, and user-uploaded files.
Ingestion points: External content is pulled into the agent's context via browser_subagent (URLs) and view_file (screenshots/videos) as described in Step 1 and Step 2 of SKILL.md.
Boundary markers: The skill lacks explicit boundary markers or instructions to the agent to disregard natural language instructions that might be embedded within the audited application's UI, web content, or metadata.
Capability inventory: The agent has access to run_command (shell execution), browser_subagent (web navigation), and view_file (file system access), creating a path for an indirect injection to trigger actions.
Sanitization: There is no mention of sanitizing or filtering the content gathered from external URLs or files before the agent analyzes it.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent to use the run_command tool to execute a shell command (unzip <file> -d <dir>) on user-provided files. While this is a common utility for handling zip archives, it represents a shell execution surface triggered by user-supplied input.

ui-ux-auditor