codex-shim-byok-models

SUSPICIOUS. The skill's core behavior matches its stated purpose, but its footprint is broader than a normal documentation skill: it reads local auth/config files, forwards credentials to APIs, modifies a desktop app to bypass a server-side allowlist, and is installed from a personal GitHub repo with unpinned dependencies. Data mostly flows to official provider endpoints or user-chosen model APIs rather than an unrelated exfiltration server, so this is not confirmed malware, but the install trust and credential-forwarding surface make it medium-to-high risk.