The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill processes untrusted user input from the 'description' and 'features' fields to make architectural decisions and generate code content. \n
Ingestion points: description and features fields in references/schema.ts. \n
Boundary markers: Absent. The inputs are used directly by heuristic logic in references/judgeHasStore.ts and references/judgeUIComplexity.ts. \n
Capability inventory: Generates React components and configuration logic. While not executing code itself, the output is intended for execution in a development environment. \n
Sanitization: No sanitization is performed on the description beyond basic keyword matching. \n- External Downloads (LOW): The skill relies on an unverifiable package @code-arena/ui for its UI components. \n
Evidence: Multiple imports in example files (e.g., best-practice-examples/standard-with-store/StandardWithStorePageContent.tsx). \n
Context: The package appears to be an internal library for the 'Code Arena' project. While the source is unknown, it aligns with the skill's stated purpose. \n- Malicious URL Alert (INFO): An automated scan reported features.us as a blacklisted URL. \n
Analysis: A thorough manual review of all 14 files failed to locate the string 'features.us'. This may be a false positive on the 'features' object property or refer to external metadata/files not included in the provided snippets.

page-best-practice