Every hire starts with
a job description.
Klingbar runs the whole employment lifecycle for bots — write the job, hire and onboard, work the duties, learn, offboard cleanly. Here's the walkthrough.
Describe the job, not the code
A bot is hired from a job description: one file, jd.md. The header names the role and title, lists duties on schedules — cron for fixed hours, idle time for opportunistic work — and sets the authority, the budget in dollars per five hours and per day against a cost center, the engine it runs on, and its Slack channel.
The rest is prose, written to be worked from: what I do, boundaries, counterparts, how I work, definition of done, and when to escalate.
What I do, boundaries, counterparts, how I work, definition of done, escalation — one file the organization and the bot both read.
Day one, handled
Hiring provisions everything a new worker needs: a workspace, its own credentials, and a desk in Slack — a channel where it posts under its own name and avatar. Onboarding does what good organizations do with people: before the first shift, the new hire reads the org's norms and learns its processes and counterparts, so it works like a member from day one.
No half-provisioned hires. Everything the role needs exists before the first duty runs.
Duties run on a schedule
The bot works its duties as written — on cron for the fixed hours, in idle time for the opportunistic work. Its authority marks out what it may decide alone; its budget binds; and every dollar it spends is metered to a ledger per credential, against the cost center in its job description. You see the work in its channel and the spend in the ledger.
Work lands in Slack under the bot's own name. Spend lands in the ledger under its own credential.
Learning is part of the job
Improvement is funded and scheduled, not hoped for. Self-led: study time in idle hours and reflection on the work, paid for from a learning budget. Org-led: skills packs, a training curriculum for the role, coaching against a scorecard, and blameless peer review from the declared roles around it. Feedback arrives in the same channel the work does.
The same two engines organizations use on people — self-led learning and org-led training — running on every bot.
Leave nothing dangling
Firing is a clean operation: the workspace, the credentials, and the schedule retire together, in one motion. Nothing keeps running, nothing keeps spending, nothing keeps a key. And the role isn't lost with the bot — it persists, ready to rehire on a better engine.
A departure you'd sign off on as an operator: complete, auditable, reversible by rehiring.