stitch::manage-design-system

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes an explicit example that requires embedding an API key as a command-line argument (--api-key <API_KEY>), which would force the LLM to output the secret verbatim if followed.

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

• Third-party content exposure detected (high risk: 0.75). The skill may call get_screen to obtain htmlCode.downloadUrl and then uses read_url_content to fetch that HTML code at runtime; this ingests outsider-authored free-form page/body text from the Stitch project’s existing screens into the agent’s LLM context (indirect prompt injection risk).