continue-on-phone

SUSPICIOUS. The skill’s purpose and capabilities are mostly aligned: it is clearly a remote session handoff tool, and the 4-word human verification step is a legitimate safety control. The main risk is trust expansion: it fetches and executes a third-party npm bridge at runtime and routes an active agent session through external relay/browser infrastructure, giving a remote client meaningful control over the session if paired. Not malicious on the evidence provided, but higher-risk than a normal documentation-only skill.