answerrank-browser

SUSPICIOUS. The workflow mostly matches the stated purpose of measuring in-browser AI answers, and it includes sensible limits like no password entry and stopping on CAPTCHAs. The main issue is trust: the skill depends on an unverifiable `answerrank` CLI and sends it collected content plus judge-model credentials, which is a high-risk credential-forwarding and supply-chain pattern even without direct evidence of malicious intent.