seo-agent

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

• Third-party content exposure detected (high risk: 0.85). The skill’s REQUIRED workflow includes runtime fetching/crawling of outsider-provided URLs and public web resources (e.g., /.well-known/mcp.json, /.well-known/llms.txt, OpenAPI specs, and developer docs) which are free-form text authored by third parties, then ingested into the agent’s LLM context during analysis.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches remote manifests at runtime (e.g., https://{target}/.well-known/mcp.json and related locations like /.well-known/openapi.json and /.well-known/llms.txt) and uses their content to drive agent tool selection and scoring, meaning external content directly influences agent prompts/behavior.