The Agent Skills Directory

[Remote Code Execution] (SAFE): No unauthorized remote execution patterns detected. The skill references standard tool-based component installation (shadcn/ui) and local build processes (mise, bun).
[Indirect Prompt Injection] (SAFE): The skill has a surface for indirect prompt injection via project-local configuration files, which is normal for its intended use case.
Ingestion points: packages/design-tokens and packages/i18n (referenced in SKILL.md).
Boundary markers: Absent, relying on the agent's internal logic for parsing structured project files.
Capability inventory: Subprocess execution for linting (bun lint) and build tasks (mise build), and tool calls for frontend component management.
Sanitization: None explicitly defined; the skill operates on the assumption of a trusted local development environment.

frontend-engineer