AI Agents for Leaders Who Have No Time to Learn AI · Section 2 · Resources

Stop explaining the same work from scratch, Complete Resource Pack

Every request, file and link from the lectures in one place. Copy a request with one click.

Arun Nagarathanam · 20 Oct 2026 Guide · 15 of 42 in this course

Lecture 2.1 · Reusable AI Skills

Why explain the same job again?

Separate what changes every week from what should survive, so the next meeting starts from what you already know instead of from zero.

Ideas to keep

01Scrolling old chats, copying fragments and introducing your colleagues again turns a helpful assistant into extra work.
02Searching past threads for the prompt that worked brings back last meeting's dates and finished items tangled with the useful instructions.
03Some details change every week: the new transcript, the status of the client's role list, the date of the next check-in.
04Other parts should survive: the Northstar background and the method for drafting blocker-focused follow-ups.
05A stronger start: Alex supplies only the new transcript and the goal, and the agent opens the saved Northstar facts and Alex's saved follow-up method.
06You stay the source of changing priorities and final decisions. The agent helps keep the records and the methods that support them.

Try it (optional)

Think of one job you repeat every week, such as the follow-up after a regular meeting.
List what changes each time (the transcript, today's status, the next date).
List what stays the same (the background of the work, the method that produces a good result).

Files and links

Northstar, Alex, Priya, Sam and Jordan continue the prepared fictional example from Section 1.

Download the Section 2 starter materials (zip)/downloads/leaders/section-02/section-02-starter-materials.zip

Lecture 2.2 · Projects and Saved Records

Why must you explain your business again?

Organize your background in 3 familiar scopes, keep 3 records inside one project, and ask the agent to show what it retrieved and what it changed.

Ideas to keep

01Use 3 familiar scopes: your company, your department or work category, and the specific project.
02A project is a folder that keeps the files, the agent's instructions and the saved records for one ongoing piece of work together.
03Inside Northstar Onboarding the agent keeps 3 records: project instructions, a file of current facts, and a decision record for each confirmed choice.
04A decision record captures a confirmed choice, the reasons behind it, and what would trigger another look.
05You do not have to write these files from scratch. Supply permitted existing notes, or explain the setup aloud, and ask the agent to draft them.
06A neat folder tree is not automatic memory. The agent uses only the background it actually retrieves, so ask it to show what it found.
07When a fact changes, the agent updates the record in place and shows you the revision.

Try it (optional)

Open practice/northstar-and-meridian-practice-files.md and read part 1: the company, department and project scopes.
Find the 3 Northstar records and the October 5 decision record's reason and revisit condition.
Open practice/skills-and-project-records.html, tab 01 Project records, and compare the before and after lines of the current facts update.
Try the two requests below with the practice file, or with records from your own work that your employer permits.

Request · Retrieve the project records and show them

Open the Northstar Onboarding project records. Retrieve the project instructions, current facts, and the latest decision record, and show me which records you found before we continue.

Use in: A ChatGPT Work task or a chat with the Northstar practice file attached

Request · Update a changed fact and show the change

Jordan has confirmed 18 of the 22 client roles for Tuesday's setup, with 4 specialist roles waiting until Thursday. Update the Northstar current facts record and show me the change.

Use in: The same task, once the records are open

Files and links

The Northstar records are prepared fictional teaching files. The retrieval readback and the update shown in the lecture are an authored illustration, not an observed agent run.

practice/northstar-and-meridian-practice-files.mdpractice/skills-and-project-records.html

Skills and project records, the prepared example: project records/resources/library/leaders/skills-and-project-records-prepared-example#records
Northstar and Meridian practice files/resources/library/leaders/northstar-and-meridian-practice-files
Open the interactive skills and project records page/downloads/leaders/section-02/skills-and-project-records.html

Lecture 2.3 · Create an AI Skill

How do you keep the method that finally worked?

Turn a conversation that finally worked into a thorough reusable prompt, test it in Work, and save it as the meeting-follow-up skill.

Ideas to keep

01Copying only the final output loses the questions, corrections and reasoning that made the draft useful.
02In ChatGPT, Work is the mode where the agent carries out a task using the files and apps you allow it to reach, rather than only replying in the chat.
03Start inside the conversation that worked, while the goal, the clarifying questions and your corrections are still visible. You do not write a template with blanks.
04A complete capture request asks ChatGPT or Gemini to write a thorough, versatile prompt that keeps the purpose, reasoning, inputs, quality criteria and rules for adapting.
05The generated prompt keeps the method and leaves this meeting's dates, such as Friday's October 2 transcript, out of the permanent instructions.
06Copying a prompt transfers your instructions, but not file access, tool connections or workspace permissions. That is why you test it in Work.
07@skill-creator is ChatGPT Work's built-in helper that turns a successful method or prompt into a saved skill. It asks a few questions first.
08Two routes: from standard ChatGPT or Gemini, generate the prompt, test it in Work, then use @skill-creator. If the conversation already happened in Work, call @skill-creator directly.

Try it (optional)

Open a conversation where AI finally produced a result you trust, in ChatGPT, Gemini or Work.
Paste the complete capture request below into that same conversation.
Read the prompt it returns: does it keep the method and leave this task's dates and names out?
If your company gives you Work, test the prompt in a Work task with permitted material, then ask @skill-creator to capture it.
If you do not have Work, keep the saved prompt and use it in an ordinary chat.

Request · Complete conversation-capture request

Act as an expert prompt engineer. Examine this successful conversation, including the original goal, the approach we used, my corrections, the useful final output and the criteria that made the result useful. Use the relevant role, business context, preferences and source material I have supplied here.

Create a comprehensive, detailed, thorough, holistic and versatile prompt that preserves this successful method and can guide another AI agent through similar work. Cover the relevant purpose, specialist perspective, usable expert methods, reasoning, inputs, output criteria, important exceptions and action boundaries. Be thorough about what matters to this job, without unrelated padding or a requirement that every prompt be long.

Separate stable instructions from changing task facts. Explain how the agent should use newly supplied material and adapt when the audience, circumstances or available information changes. Do not embed this task's temporary facts as permanent assumptions. Preserve the reason behind important choices and corrections so the method can remain useful in a different situation.

Infer information only when the conversation and supplied context support it. If an essential missing fact would materially change the method, ask me naturally before finalizing. Do not create placeholders, blank form fields or a template that makes me repeatedly rewrite my context. Ask for fresh task material when the work needs it, and retrieve relevant records only when access is available and permitted.

For substantial or unfamiliar work, have the executing agent propose the approach and ask consequential questions before execution, so I can refine the plan. For familiar small work, let the agent use an accepted method directly. Include checks that judge the result against the actual goal. Preserve appropriate expertise, privacy, permissions, decision ownership and limits on external actions; do not invent tool access or authority.

Return the complete ready-to-use prompt. Briefly explain any important choices in how you generalized the method and flag unresolved limits. Do not claim that the prompt has already been tested, that the next task will succeed or that a skill has already been created.

Use in: The successful ChatGPT, Gemini or Work conversation, at the end

Request · Capture the method directly in Work

@skill-creator Capture the meeting follow-through method from this conversation as a reusable skill named meeting-follow-up. Keep the purpose, clarifying questions, and action-and-reason rules in the skill, and leave this meeting's dates and participant facts outside the saved instructions.

Use in: ChatGPT Work, in the thread where the result was verified

Files and links

If your company does not give you Work, you can still follow every step up to the saved prompt and use that prompt in an ordinary chat. In the ChatGPT desktop app, a saved skill opens from Skills in the sidebar. The Northstar capture is a prepared illustration.

prompts/section-02-prompts.md

ChatGPThttps://chatgpt.com
Geminihttps://gemini.google.com
Skills and project records, the prepared example: skill capture/resources/library/leaders/skills-and-project-records-prepared-example#capture
Meeting Follow-Up Skill/resources/library/leaders/meeting-follow-up-skill

Lecture 2.4 · Refine an AI Skill

Why are you correcting the same thing again?

Run the saved skill on a different meeting, spot the one gap that keeps returning, and put one small rule in the skill while today's facts stay in the project record.

Ideas to keep

01A saved skill proves its value on the second meeting, not the first. A fresh transcript shows whether the method adapts to new facts.
02In Work, type @ to select meeting-follow-up and supply the new transcript. The skill supplies the method; the transcript and project record supply this week's facts.
03Much of the result works first time. The gap: Priya and Sam share the security check before Thursday's 2 p.m. session, and the draft never names who confirms readiness.
04Editing that line by hand every week keeps the supervision. Pasting Thursday's details into the skill leaves next week with outdated facts.
05Ask for the smallest change to the skill, and see the exact line before you approve it.
06Judge a refinement by the draft and by your total effort across explaining, reviewing and correcting.

Try it (optional)

Open practice/northstar-and-meridian-practice-files.md, part 2, and read Wednesday's technical check-in transcript.
Find the step Priya and Sam share before Thursday's 2 p.m. session.
Read part 3: the one rule added to meeting-follow-up, and the line kept in the Northstar project record instead.
Next time you correct the same thing twice in your own work, ask for the smallest skill change and keep the dates out of it.

Request · Ask for the smallest skill change

When 2 colleagues share a preparation dependency, the follow-up should name the single person who owns confirming completion back to me. Propose the smallest change to the meeting-follow-up skill that adds this rule, and keep Thursday's session details in the project record.

Use in: ChatGPT Work, after reviewing the drafts @meeting-follow-up produced

Files and links

The skill file in the zip is the version before this refinement. The added rule is shown in part 3 of the practice file and on tab 03 of the prepared example.

skills/meeting-follow-up/SKILL.md

Skills and project records, the prepared example: skill refinement/resources/library/leaders/skills-and-project-records-prepared-example#refinement
Meeting Follow-Up Skill/resources/library/leaders/meeting-follow-up-skill

Lecture 2.5 · Adapt an AI Skill

What if you do not know the best approach yet?

Start an unfamiliar, high-stakes evaluation from a trusted method, a role and 4 criteria, then correct the agent's plan before it runs.

Ideas to keep

01A request for proposal, or RFP, is a document a prospective client sends asking suppliers to bid on the work it describes.
02Maya must decide whether to bid on Meridian Manufacturing's full scope, propose a narrower phased scope, or decline.
03Asking AI to act as a senior strategist and tell her whether to bid gives a confident essay with no grounding in her team's real capacity.
04A stronger start combines a trusted skill or expert method with a specific role and concrete criteria for good work.
05Maya's 4 criteria: delivery requirements, specialist capacity impact, viable alternatives, and assumptions that could change the decision.
06For substantial, unfamiliar work, ask the agent to explain its approach and list missing inputs before it runs the full comparison.
07Maya spots the flaw: total headcount looks sufficient, but the integration needs a senior integration specialist, and both of hers are committed in weeks 3 and 4.

Try it (optional)

Open skills/option-comparison/SKILL.md and read 'Propose the evaluation plan before execution'.
Open practice/northstar-and-meridian-practice-files.md, part 4, for the Meridian capacity note and Maya's 4 criteria.
Give an AI tool the practice file and the first request below, then read the plan it proposes before anything runs.
Look for an assumption only you would catch, as Maya did with the specialist capacity, and correct the plan first.

Request · Ask for the evaluation plan first

Use the option-comparison method from the perspective of a commercial and delivery reviewer to evaluate the Meridian Manufacturing proposal. Define good work through 4 criteria: delivery requirements, specialist capacity impact, viable alternatives, and assumptions that could change the decision. Explain how you will evaluate this proposal against my 4 criteria and list any missing inputs you need before you run the full comparison.

Use in: ChatGPT Work or a chat with option-comparison and the Meridian practice file attached

Request · Correct the plan before it runs

Do not treat general engineering headcount as interchangeable. Check the timeline specifically against senior integration specialist availability in weeks 3 and 4, and include a phased-start option that moves integration to week 5.

Use in: The same conversation, after reading the proposed plan

Files and links

Maya and Meridian Manufacturing are a prepared fictional example. The agent's flawed plan and the corrected comparison are authored teaching material.

skills/option-comparison/SKILL.md

Option Comparison Skill/resources/library/leaders/option-comparison-skill
Skills and project records, the prepared example: adaptation/resources/library/leaders/skills-and-project-records-prepared-example#adaptation
Northstar and Meridian practice files/resources/library/leaders/northstar-and-meridian-practice-files

Lecture 2.6 · Skill Orchestration

Can one request handle the whole follow-up?

See how one workflow runs several focused skills in order, passes each one what it needs, and pauses or reports when information or permission is missing.

Ideas to keep

01The meeting work is split into focused skills, each with one clear job, so you do not start them one by one after every meeting.
02The main workflow is saved as Meetings and follow-through, and appears as meetings-and-follow-through when you type @.
03The handoffs: retrieve the Northstar records, read the transcript, check the review purpose with leader-interview, then meeting-follow-up writes the drafts.
04At the end, the notes, the drafts and the reminder proposals go onto one review page for Alex.
05Skill orchestration means arranging focused skills in sequence with explicit handoffs between them.
06A common mistake: chaining skill names and assuming each step knows facts it was never passed.
07Branches: no review purpose, the workflow pauses at leader-interview to ask. No calendar approval, the Monday 10 a.m. reminder stays Proposed only.
08Improve one focused skill and the improvement is designed to reach the workflow, as long as both are saved in the same place.

Try it (optional)

Open skills/meetings-and-follow-through/SKILL.md and read the table under 'Sequence and handoffs'.
For each row, find what goes in and what is carried forward to the next skill.
Open tab 04 of practice/skills-and-project-records.html and follow the 5 handoffs for Northstar.
Find the two branch rules: pausing when the purpose is missing, and Proposed only when calendar approval is absent.

Files and links

The workflow and its skills are authored course materials, designed to run this way. They have not been run in a live agent yet, so check how your own tool handles them before relying on them.

skills/meetings-and-follow-through/SKILL.md

Meetings and Follow-Through Workflow/resources/library/leaders/meetings-and-follow-through-workflow
Leader Interview Skill/resources/library/leaders/leader-interview-skill
Skills and project records, the prepared example: orchestration/resources/library/leaders/skills-and-project-records-prepared-example#orchestration

Lecture 2.7 · AI Skills

What have you saved that will help next time?

Name what now survives beyond one meeting: project records for the facts that change, saved skills for the methods that work.

Ideas to keep

01A project keeps the Northstar background and current facts in one place. A decision record keeps each confirmed choice with its reasons.
02Alex's Northstar Onboarding project holds the changing facts: 18 confirmed client roles, Thursday's working session, the October 5 decision record.
03Alex's saved skills hold the methods: ask what the next review needs to decide, state each action and reason, name one confirmation owner.
04When the next meeting arrives, the transcript and project record supply this week's facts and the saved skill supplies the method.
05A saved skill is a stronger starting point that improves through use, not a finished machine that never needs your review.
06Because the skill keeps the reason behind each instruction, the agent can adapt when the topic moves to a budget check or a technical rehearsal.
07A saved skill does not give the agent access to any file you have not supplied or connected.

Try it (optional)

Download the Section 2 starter materials and open README.md.
Sort one recurring job of your own into two lists: facts for a project record, and the method for a skill.

Files and links

Faster preparation brings a new pressure: more updates, options and requests competing for your attention. Section 3 takes that on.

section-02-starter-materials.zipREADME.md