AI Agent Team — Roster
Section titled “AI Agent Team — Roster”The management view of the dept-based AI agent team: which depts are chartered, which have an operating agent persona, and at what autonomy level. One row per dept — this is the org chart for KB-OS’s agent layer.
How an agent runs (the charter chain):
CONSTITUTION.md→ deptJOB_DESCRIPTION.md(policy: scope, decision authority, hand-offs) →Staff/<Role>.md(operating instructions + autonomy) → the task’s SSOT doc. As of 2026-07-09,/task-startand/task-continueauto-load this chain from the task file’s dept path — no manual loading needed.
Core (cross-business)
Section titled “Core (cross-business)”| Dept | Charter | Agent (Staff) | Autonomy | Inbox | Status |
|---|---|---|---|---|---|
| AI | JD | AI-Director | A1 | Inbox | ✅ Active |
| AI/Tools | JD | Skill-Developer (IC → AI-Director) | A1 | Inbox | ✅ Active |
| Processes | JD | Processes-Director | A1 | Inbox | ✅ Active |
| IT | JD | IT-Director | A1 | Inbox | ✅ Active |
| Cancer50 | JD | — | — | Inbox | 🔶 Dormant mission anchor |
MBR (current business focus)
Section titled “MBR (current business focus)”| Dept | Charter | Agent (Staff) | Autonomy | Inbox | Status |
|---|---|---|---|---|---|
| Strategy | JD | CEO (Talbot) | — | Inbox | ✅ Chartered |
| Offerings | JD | — | — | Inbox | Chartered, no agent yet |
| Mktg (+ Content, Viral-Growth, WealthCare50) | JD | — | — | Inbox | Chartered, no agent yet |
| IT (+ Apps, Data, Processes, Sites, UX) | JD | — | — | Inbox | Chartered, no agent yet |
| Risks | JD | — | — | Inbox | Chartered, no agent yet |
SDC / FSS
Section titled “SDC / FSS”Per-task activation — no depts chartered yet. Activate on first real need (/dept-init); do not fabricate scope for a stub dept.
Expanding the team (add an agent)
Section titled “Expanding the team (add an agent)”- Charter the dept —
/dept-initscaffoldsTasks/,Inbox.md, etc.; writeJOB_DESCRIPTION.md(scope owned, inputs/outputs, decision authority, hand-offs) from an existing one as template - Draft the Staff file —
<Dept>/Staff/<Role>-Director.md, following theCore/AI/Tools/Skill-Developer.md/*-Director.mdpattern: mission, “reads before acting” list, autonomy level, how it works a task - Set autonomy — A0 (ask everything) → A1 (draft, CEO approves) → A2 (autonomous within budget) → A3 (within scope) → A4 (within mission). New agents start A1
- Register — add the row here; Core depts also go in
~/ai-config/AGENTS.md’s dept→charter map and theCore/DASHBOARD.mdActive Depts table - Dogfood immediately — run a real task through
/task-startin the new dept’sTasks/folder; gaps show up in use, not review (KB-OS lesson, 2026-06-25)
Autonomy escalation rules
Section titled “Autonomy escalation rules”- Globally-deploying commits and changes to the
AGENTS.mdrules system always escalate to the CEO (Talbot), regardless of dept autonomy - A dept’s level lives in its
JOB_DESCRIPTION.mdfrontmatter (autonomy:); the Staff file restates it operationally - A2+ activates only when budget tracking exists — see UPGRADES (Autonomy Levels section)