fix(skills_hub): handle bytes content in bundle_content_hash

[Linux2010]

What broke

Running hermes skills update crashed with AttributeError: 'bytes' object has no attribute 'encode' when skill bundles contained binary files (e.g. audio assets from official skills like neutts).

Root cause

SkillBundle.files is defined as Dict[str, Union[str, bytes]], but bundle_content_hash() assumed all file content was str and called .encode("utf-8") on every value. When the value was already bytes, this caused a crash.

Traceback:

AttributeError: 'bytes' object has no attribute 'encode'. Did you mean: 'decode'?

Why this fix is minimal

Single function fix - detect type and route bytes directly to hash.update(), encode strings as before. No broader changes to:

• SkillBundle class or file loading logic
• Other hash computation functions
• Quarantine/installation logic

Patch (4 lines added):

content = bundle.files[rel_path]
# files can contain str or bytes; handle both
if isinstance(content, bytes):
    h.update(content)
else:
    h.update(content.encode("utf-8"))

What I tested

Added regression test verifying:

• bytes content is handled without crash
• Hash computation is deterministic (same content = same hash)
• Mixed str + bytes bundles work correctly

What I intentionally did not change

• No changes to SkillBundle class definition
• No changes to how files are downloaded or loaded
• No changes to content_hash() in skills_guard.py
• No changes to update checking or installation flow

Fixes #2739

[teknium1]

Thanks for the contribution, @Linux2010!

Closing this as a duplicate of #2740 (by @Mibayy), which targets the same fix/feature. We're consolidating on that PR for review.

If you want to help push it over the line, please jump in there — or if you think your approach is better for a specific reason that isn't covered in the other PR, let us know and we can reopen.