Why we need skills with having similar commands?

[tottishi05]

I can see that the file contents in a specific command(e.g. opsx-new.md) are very similar to that in the relative skill(e.g. openspec-new-change). After I execute the specific command like /ospx-new, when the llm needs to call the relative skill?

I think the command has already taken the complete context to the agent and llm, so there is no need to use the relative skill.

[s-KaiNet]

Skills are loaded by demand, i.e. when you use natural language "let's start a new change" llm will load the "new-change" skill and will start creating a change. With commands you are just more explicit with your intents (more predictable and user-invoked). When you use command, skill is not loaded. Skill still needed in other cases though (natural language) or if we do some integrations with other llms or agents - they will use natural language and the right skill will be activated.

[TabishB]

There's already been some discussion here, will just link it here: #611