The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from external sources, including live URLs, source code files, and images.
Ingestion points: URL input for live site audits, directory paths for codebase audits, and image files for screenshot audits.
Boundary markers: The instructions do not specify explicit delimiters or instructions to the subagents to ignore embedded commands within the analyzed content.
Capability inventory: The skill utilizes agent_browser for network navigation and subagents for automated analysis of the provided inputs.
Sanitization: No explicit sanitization or filtering of the external content is described before it is processed by the analysis tools.
[COMMAND_EXECUTION]: The skill uses local shell commands to identify the project environment and locate relevant UI files.
Evidence: SKILL.md contains bash snippets using ls, git diff, and find to check for visual changes and list source components. These commands are scoped to the local directory and used for orchestrating the audit depth.

stelow-product-ux-critique