implement-design

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

• Third-party content exposure detected (high risk: 0.85). The skill’s Step 0 “Acquire the source” can fetch a Claude Design share URL at runtime (WebFetch of https://api.anthropic.com/v1/design/...), and the returned design file’s text (e.g., labels/notes/JSX) is then ingested into the agent’s LLM context for Step 1–4, making it outsider-authored free text.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly instructs the agent to WebFetch a Claude Design share at runtime (https://api.anthropic.com/v1/design/h/?open_file=), and the fetched design file is required and directly controls what the agent implements, so this is a runtime external dependency that controls agent behavior.