Short definitions for terms that appear in the lessons. Each entry fits on one screen.
The amount of text a model can consider in one reply. Longer inputs leave less room for the answer.
An answer that reads smoothly but is wrong or unsupported. Confidence is not evidence.
A table layout you define before extraction: field name, type, definition and safe fallback values.
Steps that run on a schedule or trigger without you typing each time.
Software that can choose tools and take actions, not only write text.
The tool environment where an agent runs: folders, connectors, permissions and logs.
A structured way for one program to ask another for data or action.
A pre-built link between your workflow tool and another service.
Configured limits on what an agent can read, write or reach on the network.
How far one mistake can spread: files touched, messages sent, money moved.
Untrusted text that tries to change what the agent does next.
Repeatable under the same inputs and settings. Not the same as correct.
Running the same job twice has the same effect as running it once.
An automatic HTTP call when something happens in another system.
A durable done-list keyed by item, read before work starts.
A control that stops new automated runs. It may not stop work already in flight.