The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it searches for and interprets existing code patterns (CSS tokens, component libraries, and configuration files) to guide its design decisions. This could allow malicious instructions embedded in a project's codebase to influence the agent's output.
Ingestion points: Files searched during Layer 0 (Context Detection), such as CSS, design tokens, and component files.
Boundary markers: Absent; there are no instructions to delimit or ignore instructions within the scanned files.
Capability inventory: Use of platform search tools (Glob, Grep), browser automation (agent-browser, claude-in-chrome), and suggestion of external setup commands.
Sanitization: Absent; the content is processed directly to identify design signals without escaping or validation.
[COMMAND_EXECUTION]: The skill directs the agent to utilize platform-native search utilities like Glob and Grep for codebase exploration. It also specifies the use of browser automation tools such as Playwright, Puppeteer, or the agent-browser CLI to verify visual results.
[EXTERNAL_DOWNLOADS]: The skill includes a fallback procedure that instructs the agent to prompt the user to execute a setup command (/tunan:setup) to install the agent-browser dependency if it is not detected in the environment.

frontend-design