The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes user-provided topics and instructions to generate HTML content. While primarily a legitimate function, this creates a surface for indirect prompt injection if a user provides a topic containing malicious instructions that the agent might follow while generating the artifact. The skill includes a specific instruction to 'Escape user-provided text before placing it into HTML,' which acts as a primary mitigation against cross-site scripting (XSS) and injection within the output artifact. The ingestion point is in SKILL.md under the 'Resolve the request' and 'Create the HTML artifact' sections, where user input is interpolated into the HTML generation process. The capability inventory includes the ability to write files (HTML artifacts). The presence of the escaping instruction reduces the risk level.

eli5