mushi-integration

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

• Third-party content exposure detected (medium risk: 0.30). Stage 3 (“mushi stories map --url https://your-app.com --wait”) invokes Firecrawl to crawl and ingest free text from the caller-provided live URL, which an outsider can populate with poison content.

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

• Potentially malicious external URL detected (high risk: 0.90). The command "mushi stories map --url https://your-app.com --wait" (line 62) fetches the provided live URL at runtime and the skill then uses the crawled content with Claude to draft user stories, so external content directly controls the agent's prompts.