1. What's the purpose of the "random padding"? Remember that an LLM is incapable of thinking logically, so it'll often do things that seem purposeful but aren't.
2. Consider using an actual checksum like CRC-32 rather than a truncated hash if your intent is to detect data corruption (see https://stackoverflow.com/a/72580031)
3. Obfuscation is a fool's errand; as you said, it provides no security, but therefore it also serves no purpose except to give a false sense of security, which is often more dangerous than none at all. Actual private key container formats (like PKCS#8 and #12) allow for encrypting the key with a passphrase, which is something you might consider instead.
4. "AES-GCM keys" are not a thing; they're just AES keys (which are just random bytes of a set length).