Projects and Skills
Two features keep longer-running work organized: Projects group related conversations and context, and Skills teach LlamaBoss how you like recurring tasks done.
Projects
A project is a home for related conversations. In the sidebar, conversations group under collapsible project headers — drag a chat onto a project to move it there, or use the right-click menu. New chats started inside a project inherit its context.
You can attach files and reference material to a project so every conversation in it starts informed — useful when you keep coming back to the same codebase, report, or dataset.
Skills
A skill is a reusable set of instructions — and optionally a helper script — that tells the model how to handle a particular kind of task: your report format, your naming conventions, a multi-step routine you run every week.
Each skill lives in its own folder with a SKILL.md file describing what to do. When a conversation calls for it, the agent reads the skill and follows it, so you get consistent results without re-explaining yourself every time.
You don’t have to write the file yourself. Choosing New Skill… starts a design conversation: describe what you want the skill to do, answer a few questions, and when it sounds right say “draft this Skill” — LlamaBoss writes the SKILL.md (and helper script, if one is needed) for you.
Connections
Some skills talk to outside services — a spreadsheet platform, an email API. Connections are where their keys live: open Settings → Connections → Manage to add a provider and key. Values are stored encrypted on your machine and are never sent anywhere by LlamaBoss itself.
When a skill’s script runs, each connection is exposed to that script as an environment variable named <PROVIDER>_<KEY> — for example a smartsheet / access_token connection becomes SMARTSHEET_ACCESS_TOKEN. The script reads it with os.environ; the key never needs to appear in the skill file or the conversation.
Conversation notes
For long tasks, the agent can keep a notes file inside the conversation's workspace — recording decisions, progress, and details to carry forward. It reads its own notes back as the work continues, which keeps multi-step jobs on track.