GenAI Field Guide

Agents and beyond

Agents

An agent is a language model in a loop: read the context, decide, call a tool, read the result, repeat until a stop condition. The intelligence is the model; the ability to act is the tool set; the persistence is files and memory.

≈2 minLandscape facts checked — re-verify before relying on them.
1 · Readcontext Messages + toolresults 2 · Decide Choose the nextstep 4 · Read theresult Repeat withupdated context 3 · Call atool Act through theharness Stop when the test passes Or when the budget runs out
The loop runs until the agent decides it is done or hits a budget. "Done" is its own judgment unless you give it a test.

What makes an agent

Part What it is Who set it
The model The next-token predictor The vendor
The tools Functions it may call: search, files, calendar, code, connectors A human granted them
The loop Read, decide, act, read the result, repeat A human designed it
The stop condition "Done," a budget, an error, or a test that passes A human set it
Persistence Files, project, memory You

How to read a transcript

An agent's transcript shows every layer. Roughly, top to bottom:

Line in the transcript Who wrote it
System prompt The vendor and your organization
Tool definitions (search, files, each connector, each skill) The harness
"Draft the SITREP from today's reports." You
Tool call: read_file("reports/0908.txt") The model, as a structured request
Tool result: the file's text (possibly truncated) The harness
The draft The model
"[Earlier turns summarized]" The harness, on compaction

If there is no tool call, there was no search and no file read. That is how you tell "it looked it up" from "it answered from training."

Try it (needs a terminal agent or a builder such as Agent Designer, if your role permits)

Write a thirty-line skill for "how our office formats a weekly SITREP." Load it. Compare the output before and after. Then give an agent a task that needs a tool it lacks, add the tool, and watch the behavior change.