The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests external and potentially untrusted data—including the content of the target URL, visual screenshots, and user-provided persona descriptions—and interpolates these values into the instructions for sibling skills. This presents a surface for indirect prompt injection, where an attacker could theoretically place malicious instructions on a website being reviewed to influence the agent's behavior.
[COMMAND_EXECUTION]: Browser automation tools such as ctx:playwright and webapp-testing are employed to interact with remote URLs and capture visual state, which is a standard part of its intended functionality for UI auditing.
[SAFE]: The orchestrator is designed to avoid autonomous execution; it requires the user to manually invoke each lens and confirms progress through approval gates. This architecture ensures that the agent does not perform sensitive operations or skill chaining without explicit user oversight.

ui-ux-review-suite