codex-chatgpt-control-bridge

SUSPICIOUS: the skill’s core behavior is mostly consistent with its stated purpose, but the trust model is weak. The main concerns are third-party transitive plugin installation from a personal GitHub repo on a mutable branch, plus a broad browser/file bridge into a signed-in ChatGPT session. I found no clear evidence of credential theft or covert exfiltration, so this is better classified as a high-trust, medium-risk integration rather than confirmed malware.