Osysharp.Agents.Teaching — the two screens people teach an agent on

`Osysharp.Agents.Teaching` draws "Correct this" on a piece of an agent's work, and an agent's Instructions page: what is waiting to be read, what is active, what was retired, who decided, and every change as a numbered set that can be rolled back.

You are here: Building with agents › The kits › Agents.Teaching

Summary#

Osysharp.Agents.Teaching draws "Correct this" on a piece of an agent's work, and an agent's Instructions page: what is waiting to be read, what is active, what was retired, who decided, and every change as a numbered set that can be rolled back.

The screens decide nothing. Who may correct, who may make an instruction active and who may roll back are answered by Osysharp.Agents. A control a person may not use is not offered.


Package	`Osysharp.Agents.Teaching`, version 1.2.0
You add it with	`use Osysharp.Agents.Teaching;` in the app's manifest.
It sits on	[[agent-kit-agents\|Agents]]
It also uses	`Identity`, `Ui` — parts of the platform's own library
Its own test app	A small maritime museum whose labels are written by an AI agent — `kits/agents-teaching/tests`

Description#

What do you get without writing anything?#

A correction form: what was wrong, what it should have been, why — and an instruction proposed from it in the same breath.
The Instructions page for an agent: propose, edit, activate, reject, retire, roll back.
For each active instruction: how much work has been sent it, and whether that work was corrected since.

When do you need it?#

When people who are not developers should improve an agent. The functions underneath (CorrectAgentWork, ProposeAgentInstruction, ActivateAgentInstruction) are in the core kit and work without these screens.

What does it expose?#

The tables below are generated from the kit itself, not written by hand: osy kits Osysharp.Agents.Teaching --surface lists each of them with the code of the kit's own test app that uses it, and the tests that show it working.

Functions you call, those its own tests show most first.

Function	What it does	Tests that show it
`InstructionBoardOf(string agent)`	An agent's instructions, for its page: what is waiting to be read, what is active, what is past, and every change to the set — each as whoever is looking may read it.	6
`SendCorrection(Guid task, CorrectionForm form)`	Send a correction from the screen — and, when the person asked, propose an instruction from it in the same unit of work.	5
`WorkTeachingOf(Guid task)`	One piece of work, for teaching: what it was told, what it handed back, each step of its record, and the corrections people have made of it.	4
`ActivateFromPage(Guid instruction, string why = …)`	Make a proposed or retired instruction active.	4

Screens and components you mount — each is one line in a page of yours.

Component	What it draws
`CorrectThis(task, step, proposal, label, agent, changed, quiet)`	"correct this".
`CorrectionLine(c)`	One correction, as it reads on the work it corrects.
`WorkTeaching(task, beside)`	Teaching, on one piece of work — what a run's page shows so that a person can correct it: what it was told when it ran (and what the agent has been told since), what it handed back, each step of its record, each with "Correct this"; and the corrections made so far.
`AgentInstructions(agent)`	An agent's instructions, as a page's body.
`InstructionRow(card, mayTeach, agent, changed)`	One instruction: what the agent is told, what work it is for, where it came from, who decided — and, for whoever teaches, what can be done with it.
`SetRow(set, agent, agentName, mayTeach, changed)`	One line of the history: a change to the set, and the way back to it.

Go deeper#

Correct this, and an agent's Instructions page — the screens people teach an agent on (Agents.Teaching kit) — the two screens
Correcting an agent, and what it is told from then on — corrections, instructions, instruction sets and rollback (Agents kit) — corrections, instructions, instruction sets, rollback