ROLE
What the system should do.
METHOD / BUILDING SYSTEMS
An agent is not a prompt.
A model is only one component. An agent appears when role, memory, tools, rules, permissions, verification and actions surround it.
CONTEXT
What it needs to know.
MEMORY
What it retains.
TOOLS
What it can reach.
RULES
Boundaries and permissions.
REASONING
Where a model helps and code decides.
VERIFICATION
Tests, logs, security and humans.
ACTION
What it may change.
Intelligence becomes useful when it enters a system.
HUMAN IN THE LOOP
AIproposes → checks → executes a low-risk action
AIdetects uncertainty → HUMAN → confirms a high-risk action