The Agent Skills Directory

[SAFE]: The skill is primarily composed of architectural guidelines, React code templates, and decision-making logic for frontend development. It operates by generating standardized code based on local references and does not attempt to execute dangerous system commands or exfiltrate data.
[INDIRECT_PROMPT_INJECTION]: The skill ingests a user-provided description field to guide the generation process and determine UI complexity. This creates a surface where a user might attempt to inject instructions to override the agent's behavior or influence the generated code. However, the skill lacks high-privilege capabilities (like file system writes outside the project, network access, or shell execution) that would make this an exploitable vulnerability in the context of the skill's intended use.
Ingestion points: The description field in the PageGeneratorSchema (defined in references/schema.ts) and user-provided input processed according to the instructions in SKILL.md.
Boundary markers: The skill does not implement specific delimiters or 'ignore embedded instructions' warnings for the description input during its interpolation into the generation context.
Capability inventory: The skill is restricted to code generation tasks; no subprocess calls, network exfiltration patterns, or dangerous dynamic execution methods were found across the provided files.
Sanitization: The input is validated against a Zod schema for structural integrity, but no content-level sanitization or instruction filtering is performed on the descriptive text.

page-best-practice