| name | country-economy-jtbd-index |
| description | Index and shared framework for the Country-Economy Jobs-To-Be-Done skill library: national operating systems, their AI-personnel role skills, work-system completeness roles, 15 cross-cutting archetypes, and the AI/robot catalogs. Use this skill first to navigate the library, understand the shared teaming pattern and accountability model, and find the right operating-system or role skill for any economic task. |
Country-Economy JTBD Skill Library — Framework & Index
This library turns the document Country-Economy Core Jobs To Be Done into deployable skills. It is organized so an LLM or agent can find the right context for any job in a modern economy, understand the human/AI/robot division of labor, and respect the human-accountability boundaries.
How the library is organized
00-framework/ — this index plus the shared concepts every skill assumes (you are here).
01-… through 22-… — one folder per national operating system. Each has a sector SKILL.md (orchestrator) and a roles/ subfolder of AI-personnel role skills.
strategic-missions/ — cross-cutting national missions (energy abundance, semiconductor sovereignty, bioeconomy, frontier-AI production, quantum & space, strategic supply chain, science-to-industry, talent formation, public procurement, cyber defense, advanced manufacturing, digital infrastructure). A mission is an orthogonal axis to the sectors: it composes several of them toward one objective.
cross-cutting-archetypes/ — the 15 role patterns (Strategist, Operator, Builder, …) that recur in every sector.
_catalogs/human-command/ — accountable human owners for the strategic missions and cross-cutting authority (national technology strategist, AI governance lead, import/export compliance lead, procurement innovation lead).
_catalogs/informal-economy/ — support roles for the informal and subsistence sector (the majority of employment in much of the world): vendor support, gig/platform coordination, informal transport, waste-picker cooperatives, smallholder advisory, savings groups, mutual aid, and formalization navigation — designed to strengthen, not surveil, informal workers.
_catalogs/ai-personnel/ and _catalogs/humanoid-robots/ — reusable cross-economy role patterns.
_catalogs/autonomous-machines/ — non-humanoid autonomous platforms: self-driving cars/trucks/shuttles, autonomous tractors and harvesters, loaders and earthmovers, mining haul trucks, drones (survey, spray, delivery), warehouse movers, and surface vessels. Several sectors also nest domain-specific machines under <sector>/autonomous/ (e.g. 05-food/autonomous/, 11-transportation/autonomous/, 08-mining/autonomous/).
_catalogs/embodied-ai-stack/ — the roles that build and operate both the LLM-brained robots and the autonomous machines: brain/autonomy orchestrator, VLA policy engineer, world-model engineer, robot-gym/sim-to-real engineer, RLAIF pipeline engineer, evaluation/red-team agent, fleet safety officer, teleoperation operator, fleet operations agent, and data/telemetry engineer.
_catalogs/autonomous-fleet-ops/ — the operations layer for autonomous vehicle/machine fleets: ODD & safety-case engineer, remote-operations (teleop) center supervisor, HD mapping & localization engineer, V2X/connectivity & infrastructure engineer, homologation & regulatory lead, depot/maintenance lead, in-field safety operator, and incident/disengagement analyst.
_catalogs/capability-optimization/ — the how-it's-built layer: the model tiers (LLM, SLM, tiny LM, deterministic) and the spectrum of optimization methods (imitation, model-based/offline RL, RLHF/RLAIF, sim-to-real, distillation/compression, classical control, search, formal methods) with the roles that select and run them. RLAIF is one option among many.
_catalogs/simulation-training/ — the anti-deskilling / keep-warm layer: job and role simulators that keep humans current, rebuild the learning ladder, and capture tacit knowledge — reusing the machine-training world models. Curriculum designer, scenario-generation agent, competency/certification agent, drill & exercise coordinator, dual-use world-model/fidelity engineer, and tacit-knowledge capture agent. See docs/role-simulation-and-keepwarm.md.
_catalogs/transition/ — the transition layer: the jobs that move a nation from today's economy to a human-AI-robot one. The end-state map is silent on the path, but the path is its own major piece of work with its own jobs, frictions, and failure modes — and it differs sharply by context. Roadmap architect, automation sequencing analyst, adoption & change-management agent, workforce-transition program manager, distributional-impact & gains-allocation analyst, vendor-pluralism & lock-in analyst, reversibility & rollback-readiness officer, and staged-rollout & blast-radius assurance agent. The accountable owner is the National AI-robot transition director in _catalogs/human-command/. See the Transition Dynamics and Sequencing and Political Economy of Automation sections of docs/country-economy-core-jtbd.md.
_catalogs/enabling-work/ — the work-system completeness layer: reusable ancillary roles that surround core JTBD with Enable, Integrate, Assure, Adapt, and Sustain work. Use it for research and evidence, knowledge and tools, dependencies and handoffs, expediting, decision preparation, stakeholder alignment, independent review and challenge, brainstorming, continuous improvement, administration, and sustainable workforce capacity. See docs/work-system-completeness-map.md, checklists/work-system-completeness-checklist.md, and tools/work-system-mapper.html.
The shared model every skill assumes
A job is a durable outcome society must reliably produce. A role is one way to own, coordinate, or execute it. AI personnel and robots occupy portions of roles; legal, moral, and political accountability stays with humans and institutions.
Core work is necessary but not sufficient. For each core JTBD, check the orthogonal work-system families: Enable (evidence, knowledge, tools), Integrate (dependencies, flow, stakeholders), Assure (quality, risk, challenge), Adapt (options, learning, improvement), and Sustain (administration, capacity, continuity). Do not instantiate every support role automatically; choose embedded, shared, platform, federated, or temporary support according to demand, specialization, consequence, and coordination cost. The reusable skills live in _catalogs/enabling-work/.
The universal seven-step lifecycle (used in every skill):
- Sense reality — gather data, observe conditions, inspect sources, listen to people.
- Interpret reality — diagnose, forecast, model risk, prioritize.
- Decide — choose policy, design, action, allocation, escalation, or tradeoff.
- Mobilize — assign labor, budget, materials, rights, permissions, logistics, schedule.
- Execute — perform the work in digital or physical space.
- Verify — test, audit, measure, inspect, certify, and learn.
- Govern — maintain legitimacy, safety, accountability, continuity, and trust.
The five-layer role design pattern (used to staff every job):
- Human owner — accountable for goals, values, exceptions, relationships, signoff.
- AI personnel — research, draft, analyze, monitor, simulate, coordinate, document.
- Robot personnel — fetch, carry, inspect, clean, assemble, assist, enter hazardous spaces.
- Control layer — permissions, audit logs, escalation thresholds, incident reporting, evaluation.
- Public trust layer — explainability, appeal, privacy, bias testing, safety certification, labor impact.
How robot personnel are built (assumed architecture). Robot roles in this library are LLM-brained embodied agents: a multimodal LLM brain perceives, plans, and issues physical actions as tool calls (e.g. grasp, navigate_to, place), which are executed by Vision-Language-Action (VLA) policies trained on world models (learned physics simulators), robot gyms (massively parallel sim-to-real), and RLAIF (reinforcement learning from AI feedback). Fleets may share one brain model or mix specialized ones (a deliberative orchestrator over fast reactive controllers). A verified low-level safety layer can refuse or override unsafe tool calls independently of the brain. The roles that build and operate this stack live in _catalogs/embodied-ai-stack/. A concrete, buildable instance of this exact architecture — a typed MotionIntent tool-call schema, a layered control stack where authority to stop runs opposite to authority to plan, and a verified safety state machine the brain cannot bypass — is documented in docs/embodied-reference-build.md. The same brain-and-tool-calls model extends to non-humanoid autonomous machines (vehicles, farm equipment, loaders, drones), which add an Operational Design Domain, SAE levels, a verified safe-stop, and a teleoperation fallback (_catalogs/autonomous-machines/, _catalogs/autonomous-fleet-ops/).
Capability is right-sized, not one-size — and RLAIF is one method among many. The brain need not be a single large model trained one way. Capabilities are spread across model tiers — LLM, SLM, tiny LM, and deterministic controllers — and built with a spectrum of methods: imitation/behavior cloning, model-based and offline RL, RLHF/RLAIF, sim-to-real, self-supervised pretraining, supervised fine-tuning, distillation and compression, search/planning, classical optimization and control, and formal verification. Each capability is assigned to the smallest, most deterministic tier and the most efficient method that meets its accuracy, latency, and safety bar — with a verified deterministic safety layer beneath anything learned. The roles that select and run this spectrum live in _catalogs/capability-optimization/.
Guarding against deskilling. Automating routine work erodes the human fallback bench, tacit judgment, and the learning ladder. Every sector skill carries a Deskilling watch & keep-warm section (its specific risk, countermeasures, and a job/role-simulator regime), OS 22 (Resilience) owns the cross-sector drill program, and _catalogs/simulation-training/ holds the roles that run it. The key idea: the world models and simulators built to train the machines double as the keep-warm simulators that keep humans current and rebuild the learning ladder — one simulation substrate, two students. See docs/role-simulation-and-keepwarm.md.
The command & cadence model (how delegation actually runs)
The five-layer pattern says who is on the team; this says how they run together without losing accountability. Every role and mission assumes it.
Three-layer workforce. Human command owns accountable judgment, authority, trust, ethics, and signoff (and must never lose ownership, legitimacy, escalation, redress). AI personnel own research, drafting, coding, monitoring, simulation, and coordination (and must never lose evidence, uncertainty, constraints, logs). Robot/machine personnel own bounded physical execution (and must never lose the safety envelope, human override, physical proof).
The operating loop (run it for any delegated work):
- Mission assignment — the human owner sets objective, constraints, success criteria, and risk tier.
- Context loading — agents load approved data, policies, tools, maps, and current state.
- Task decomposition — separate research, planning, execution, monitoring, verification.
- Delegation — AI does cognitive work; robots/machines do approved physical work; humans hold judgment and exceptions.
- Verification — check outputs against metrics, evidence, tests, inspections, and human-review thresholds.
- Escalation — uncertainty, rights impact, safety risk, conflict, or policy ambiguity triggers human command.
- Learning — incidents, failures, and successful patterns update SOPs, evals, prompts, maps, and training.
Delegation rules. Delegate to AI when the work is text, code, data, classification, monitoring, forecasting, simulation, routing, or first-draft synthesis. Delegate to robots/machines when it is fetch, carry, inspect, clean, sort, stage, load, unload, scan, guide, or repeatable manipulation in a bounded environment. Keep with humans when it involves force, rights, consent, accountability, public legitimacy, final professional signoff, scarce-resource triage, or unresolved ethical tradeoffs.
Required control surfaces: role charter, context pack, tool permissions, evidence log, evaluation, incident path, review cadence.
Command cadence: real-time (safety, incidents, outages, cyber, public-safety escalations); daily (queues, uptime, throughput, exceptions); weekly (metrics, quality drift, cost, adoption, workforce impact); monthly (risk register, eval results, audits, policy); quarterly (role redesign, procurement, capacity, training, public trust, resilience).
Three failure modes to design against: automation without an accountable owner; AI output treated as a final decision; a robot's task envelope expanding informally. (See checklists/ for the deployment gates and templates/ for role/agent/robot briefs.)
Universal, not US-specific. The jobs are invariant across nations; ownership, formality, and capacity are local variables. Every skill carries a "context modifiers" section so it can be adapted to any nation — any size, geography, income level, or political system.
The national operating systems
| # | Operating system | Role skills |
|---|
| 01 | Governance, Law, and Public Administration | 6 AI roles |
| 02 | Public Finance, Tax, Treasury, and Procurement | 12 AI roles |
| 03 | Defense, Intelligence, Border, and Foreign Affairs | 8 AI roles |
| 04 | Public Safety, Justice Operations, and Emergency Response | 9 AI roles |
| 05 | Food, Agriculture, Fisheries, and Nutrition | 14 AI roles |
| 06 | Water, Sanitation, and Public Hygiene | 6 AI roles |
| 07 | Energy, Utilities, and Grid Operations | 8 AI roles |
| 08 | Mining, Materials, Chemicals, and Industrial Inputs | 6 AI roles |
| 09 | Manufacturing and Industrial Production | 8 AI roles |
| 10 | Shelter, Construction, Land, and the Built Environment | 11 AI roles |
| 11 | Transportation, Logistics, Postal, and Mobility | 9 AI roles |
| 12 | Communications, Software, Cybersecurity, and Digital Infrastructure | 12 AI roles |
| 13 | Healthcare, Public Health, and Biomedical Systems | 9 AI roles |
| 14 | |
The 12 strategic missions (the other axis)
Missions are cross-cutting national capabilities that compose several sectors toward one objective. Use them when the goal is a capability rather than a sector.
| Strategic mission | Composes operating systems |
|---|
| Energy Abundance | 07, 16, 10, 11, 08, 22 |
| Semiconductor Sovereignty | 08, 09, 15, 12, 03, 11 |
| Bioeconomy | 13, 05, 15, 08, 19, 03 |
| Frontier AI Production | 12, 15, 07, 08, 20 |
| Quantum and Space Systems | 15, 08, 09, 03, 12 |
| Strategic Supply Chain | 11, 08, 16, 03, 17, 22 |
| Science-to-Industry | 15, 09, 02, 14, 16 |
| Frontier Talent Formation | 14, 20, 15 |
| Public Procurement for Frontier Technology | 02, 01, 15 |
| Cyber Defense | 12, 03, 22, 04 |
| Advanced Manufacturing | 09, 08, 11, 20, 15 |
| Digital Infrastructure | 12, 07, 16, 01, 23 |
How to use this library
- Start here to orient.
- Open the operating-system skill for the relevant sector to get the mission, JTBD, roster, and accountability boundary.
- Deploy the specific role skill(s) under that sector's
roles/ for execution, or an archetype/catalog skill for a cross-sector pattern.
- Run the work-system completeness check: which Enable, Integrate, Assure, Adapt, and Sustain functions are actually required, and should each be embedded, shared, platform-based, federated, or temporary?
- Always run the seven-step lifecycle and stop at the human-accountability boundary.
Deployment order (high-leverage first)
- Back-office document work (permits, benefits, procurement, compliance, finance ops).
- Monitoring and triage (cyber, infrastructure telemetry, health surveillance, fraud).
- Customer/citizen service (intake, routing, status, routine support).
- Planning and simulation (budgets, logistics, energy load, disaster scenarios).
- Software and data infrastructure (coding, test, data-quality, analytics agents).
- Physical logistics (warehouses, hospitals, hotels, labs, factories, facilities).
- Inspection and maintenance (utilities, plants, buildings, roads, farms, sites).
- Care support (reduce burden around care; do not replace caregivers).
- Hazardous response (robots first into dangerous, dirty, dull, degraded environments).
Work that should stay human-led (applies across all skills)
Coercive state power; rights-impacting decisions; intimate human care; democratic legitimacy; high-consequence safety; ethical and social tradeoffs; and final accountability for AI deployment, model-risk acceptance, incident response, and redress.