The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by analyzing user-provided web products, components, and interface descriptions, which establishes an inherent surface for indirect prompt injection if the ingested data contains malicious instructions.
Ingestion points: Target files, screens, or component code provided by the user as specified in the argument-hint and the initial "Scope the surface" step of the SKILL.md workflow.
Boundary markers: The instructions do not explicitly define delimiters (such as triple backticks or XML tags) or provide "ignore embedded instructions" warnings for the agent to use when processing external content.
Capability inventory: The skill involves reviewing code and applying framework-specific fixes across React, Next.js, Angular, and Vue. These tasks require the agent to have read and write access to the project's filesystem.
Sanitization: While the guidelines in references/framework-patterns.md specifically recommend sanitizing v-html and rich text within the code being reviewed, there is no mention of sanitizing or escaping the user's input before it is processed by the agent itself.
[SAFE]: The skill's content is educational and instructional, focusing on accessibility best practices like focus management, semantic HTML, and ARIA usage. All external links point to the author's own GitHub profile (aktsmm) or official Creative Commons license documentation, which are consistent with the skill's metadata and stated purpose.

web-accessibility