add-provider-models-to-pi

Not overtly malicious (no code execution, no network exfiltration, no destructive actions), but it has strong confidentiality risks: it persistently logs provider headers and payloads and also logs the full raw message content without redaction, while the destination path is controlled by an environment variable and the log is truncated/overwritten at startup. This can leak credentials/secrets (including Authorization if present) to disk and enables writing to unintended filesystem paths.