Why add an agent skill to a CLI that already has a context command?

Introduction logbasset, my open source CLI for querying Scalyr logs, was already built to be friendly to AI coding agents. It ships a context command that prints an agent-facing reference — every command, flag, output format, time syntax, and even a list of flags that look plausible but don’t exist — plus a schema command that emits the same information as machine-readable JSON. So when I sat down to add an installable agent skill, I had a fair question to answer first: if the binary already documents itself, what’s the skill even for? Self-documentation isn’t discovery The context command answers how to use logbasset. But it only helps an agent that has already decided to reach for logbasset in the first place. Nothing tells the agent “this tool exists, and it’s the right one when the user mentions Scalyr logs.” That’s a separate problem — discovery — and a context command can’t solve it, because the agent has to know to run it. An agent skill fills exactly that gap. A skill is…