do-interceptor

SUSPICIOUS. The skill's capabilities mostly match its purpose, and the flagged command/data findings are largely documentation artifacts, not direct malware signals. However, it grants very broad real-session browser and macOS control, includes an unauthenticated local bridge, mentions loading another skill, and retains extensive operational instructions despite claiming the runtime is unavailable; this makes it high-impact and medium security risk, but not confirmed malicious.

No direct malware or data-exfiltration code is present in the supplied documentation. The workflow installs a persistent, highly privileged macOS bridge and explicitly documents an unauthenticated local socket, creating a significant security risk if any same-user process or supply-chain artifact is compromised. Sparkle auto-updates are another unverified trust boundary. Review the actual bridge, update configuration, build scripts, dependency lockfile, and package provenance before deployment.

The code is a readable browser-launch wrapper, not evident malware. It deliberately opens an authenticated Chrome profile and therefore creates a significant isolation and credential-exposure risk if the URL, browser executable, profile configuration, or Chrome extensions are attacker-controlled. The unvalidated USER_PREFS sourcing is the clearest code-execution risk. Restrict USER_PREFS to a trusted, root-owned or user-owned canonical file and use a genuinely isolated browser data directory for untrusted testing.