How do you teach a professional to build a working environment rather than memorise prompting techniques?

Who the project is for

Academic Agent Workspace is an original course and working environment for university educators, researchers, learning designers, and programme leaders. Participants do not learn merely to “talk to AI”; they build a workspace that preserves the task, sources, criteria, and decision history.

The programme spans 36 academic hours and nine sessions. Every session ends with a working artifact inside the participant's real project.

Teaching problem

A one-off chat loses context, obscures claim provenance, and is difficult to hand over to a colleague. Slide-based training may teach tool names but leave no process that still works the next day.

The course moves participants from tool operator to designer of their own environment without requiring prior programming experience.

My role

• programme author and instructor;
• exercise, rubric, and quality-criteria design;
• architecture of the agent workspace and protected platform;
• project mentoring through the cold handoff.

Nine sessions, one workspace

Three modules move from workflow foundations through research and teaching practice to independent agent design. The repository grows with the participant's capability: TASK.md, source registers, claim maps, rubrics, AGENTS.md, a custom skill, and an AI-use policy are not examples—they become parts of the participant's environment.

Final artifacts

Participants finish with a connected set of outputs:

• a working agent environment and task queue;
• an evidence-based literature review;
• a lecture or presentation with reproducible source material;
• a custom skill or practicum;
• an AI-use policy and evaluation criteria;
• a mirror artifact and cold-handoff report.

Mentoring, not demonstration

An agent may prepare analysis, identify a gap, or propose questions. The educator remains accountable for pedagogical decisions, assessment, and source interpretation. Mentoring helps participants define boundaries, conduct review, and bring their own project to a state another person can take over.

How quality is assessed

Assessment is not based on how impressive a conversation looks. It checks artifact completeness, evidence use, transparent AI use, and whether a second person can run the project without oral context. A private-repository review confirmed that a build-time registry validates the material corpus and blocks incomplete publication.

What was built

A multilingual public course overview is available. A private-repository review confirmed the protected environment, search, and generated course packages; those elements are not presented as public artifacts. The project demonstrates both engineering and mentoring: translating a method into a programme, exercises, and participant autonomy.

Limitations

Participant data and cohort outcomes are not published. Learning outcomes, current product status, and the formal offer format require owner confirmation. The protected login proves that the environment exists; it is not a public demo of its contents.