[paul-v-snet]

Description

The main idea of these changes is to split skills/workflows into two types:

1. Core skills, i.e. verified and included in the core
2. Custom skills added/created by users

The core skills still remain in plugins/workflow/instructions, and the custom skills are now located in memory/workflow/skills.

The main reason for keeping the core skills in the instructions directory is that if we move them to the memory directory, they will also be included in the omega-memory volume. Because of this, they won't be updated automatically on the user's side, since they already exist in the volume and will be skipped.

So, to ensure that core skill updates are always applied on the user's side, I decided to keep them outside memory.

In addition, I defined the "workflow memory directory" as a space that will serve to store workflows/skills metadata (results, some intermediate data, and so on). This will be located in memory/workflow/space and will also be preserved in the omega-memory volume.

Regarding the memory export mechanism: I opened a draft PR in the import-knowledge-package repo that introduces a new skills component in the memory_portability module, which will now work not only with history and LTM but also with the custom skills introduced here.

According to these changes, I also had to replace the both option with the all option because now we have three components instead of two, so both would be confusing here.

I also implemented all the necessary changes on the core side, so once import-kb version 0.2.6 is released, both implementations will work together and support the history, ltm, skills, and all options.

How Has This Been Tested?

Tested locally with a live agent: added skills persisted after container restarts and image rebuilds.
Also, all memory-export features work fine with import-kb version 0.2.6.

Checklist

• PR contains autogenerated code
• Self-review completed
• Test scenarios above are passed with the version of the code from PR