The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to review and analyze user-provided HTML, templates, and components to validate breadcrumb accessibility. This creates a surface where malicious instructions could be embedded in the code being reviewed to influence the agent's behavior.
Ingestion points: The 'Code Review' section in SKILL.md specifies reviewing templates, server-rendered HTML, and shared components.
Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded within the user-provided code samples.
Capability inventory: The skill's capabilities are limited to analyzing, flagging violations, and explaining accessibility improvements.
Sanitization: There is no mention of sanitizing or escaping the content of the HTML or templates before the agent processes them.
[DYNAMIC_EXECUTION]: The React code example provided for SEO structured data in references/rule.md demonstrates the use of dangerouslySetInnerHTML to inject JSON-LD into the document.
Evidence: <script type="application/ld+json" dangerouslySetInnerHTML={{ __html: JSON.stringify(schema) }} />
Analysis: While this is a standard method for implementing JSON-LD in React applications, it is a sensitive pattern. If the schema object incorporates unsanitized user input, it could lead to Cross-Site Scripting (XSS) vulnerabilities in the user's application.

breadcrumb-navigation