office-hours

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly instructs the agent to include "verbatim user quotes" in a prompt sent to Codex and to display full Codex output verbatim, which will cause any user-supplied secrets (API keys, tokens, passwords) to be copied and relayed verbatim — an exfiltration risk.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill's setup flow downloads and executes a remote installer at runtime (curl "https://bun.sh/install" → bash "$tmpfile"), so the URL https://bun.sh/install is fetched during execution and runs remote code.