compose-page-blocks

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill explicitly requires reproducing "every visible word" from the source HTML VERBATIM and to emit source URLs, so if those inputs contain secrets or tokens the LLM is forced to include them verbatim in its output (exfiltration risk).

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). The required workflow reads the per-invocation sanitized rendered source HTML file (htmlPath) and screenshot/design foundation paths, and the model must copy text verbatim from that source HTML into post_content; this HTML is derived from outside web page content, so its free text (even sanitized) can be fed into the LLM context via the HTML-reading step, creating an indirect prompt-injection risk.