design-ship

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly imports a handoff bundle from a claude.ai/design URL (see "One command: Claude Design handoff URL → reviewable GitHub PR" and Phase 1 — Import), and the agent parses and acts on that user-provided bundle to scaffold components, generate tests, verify diffs, and open PRs, so untrusted third-party content can materially influence tool behavior.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly accepts and uses a runtime Claude Design handoff URL (e.g., https://claude.ai/design/abc123) as the required bundle input which is fetched and then drives component scaffolding and downstream agent prompts/code/test generation, so the external content directly controls agent behavior.