A portable six-layer operating system for reliable AI-assisted business work.
Operator OS gives humans and AI agents one shared, inspectable structure for process, execution, knowledge, and memory. It is based on an operating system used in a real business, generalized here without company data or private rules.
LLMs are probabilistic; many business processes require consistency. Operator OS separates judgment from repeatable execution so teams do not have to re-explain their standards every session or bury every rule in one enormous prompt.
| Layer | Purpose | Location |
|---|---|---|
| 1. SOPs | Durable process and governance | Operator Team OS/1. SOPs/ |
| 2. Agents | Optional specialist roles | Operator Team OS/2. Agents/ |
| 3. Skills | Task instructions and deterministic tools | Operator Team OS/3. Skills/ |
| 4. Workflows | Short routers and repeatable sequences | Operator Team OS/4. Workflows/ |
| 5. Knowledge | Permissioned business data | Drive - * folders |
| 6. Memory | Active context, durable facts, session handoffs | Operator Team OS/6. Memory/ |
The numbered 5. Implementation Plans/ folder tracks approved changes to the OS;
business knowledge stays outside the OS container so permissions remain clear.
- One canonical policy file with portable, non-symlink discovery pointers
- Query-first memory retrieval instead of loading whole memory files
- Atomic session handoffs instead of shared daily-log appends
- Human approval before durable facts enter Long-Term memory
- A knowledge-graph schema, master WIKI, and audit/intake workflows
- A read-only workspace doctor and lightweight context-audit skill
- Stronger privacy, secret, dependency, cache, and cloud-sync hygiene
git clone https://github.com/rangerrick337/operator-os.git
cd operator-os
python3 "Operator Team OS/3. Skills/workspace-doctor/scripts/operator_doctor.py"Then ask your AI tool: “Read AGENTS.md and help me customize Operator OS.”
Useful entrypoints:
Operator Team OS/WIKI.md— map of the system/start— begin with compact context/memory— retrieve, save, review, intake, or audit context/wrap-up— surface confidence gaps and blind spotsSETUP.md— customize the template safely
memory-read— heading-level, source-linked local retrievalmemory-manage— Active updates, proposals, and session handoffsworkspace-doctor— read-only structural and portability checksai-context-optimizer— detect duplicated or stale AI-facing contextwrap-up— end-of-session uncertainty and blind-spot review- Document, presentation, spreadsheet, and conversion examples from the original public release remain available for teams that need artifact workflows.
- One canonical source; small platform pointers
- Progressive disclosure; load only what the task needs
- Deterministic scripts for repeatable mechanics
- Explicit permission boundaries for knowledge and memory
- Evidence before synthesis; human approval at consequential write boundaries
- Reversible maintenance and inspectable Markdown
Operator OS is platform-agnostic. Different tools may need small discovery
wrappers, but canonical policy, skills, and workflows stay under Operator Team OS/.
MIT. See LICENSE and retain any additional license files shipped with bundled
third-party skills.