The five questions from Lesson 2 are the framework. This lesson teaches the execution: how to brief the expert so the interview produces operational knowledge rather than performed expertise, how to take notes that capture specific rather than generic material, how to follow the conversation where the expert's experience leads, and how to write the north star summary that anchors the first SKILL.md draft.
A single interview of sixty to ninety minutes, conducted well, produces enough material to write a substantive first-draft SKILL.md. Conducted poorly: with a defensive expert, unfocused notes, and no immediate synthesis: the same ninety minutes produces a transcript of general statements that are too vague to become agent instructions. The difference is not the expert's knowledge. It is the extraction technique.
This lesson is practical. By the end of it, you should be able to sit down with a domain expert (or with yourself, as your own first test subject) and conduct an extraction interview that produces the raw material for a SKILL.md that encodes something worth encoding.
The interview works best when the expert knows in advance what it is for. Brief them on the Knowledge Extraction Method before you begin. The briefing has three elements, and each one shifts the expert's orientation in a way that improves the quality of the material they produce.
Explain the purpose. You are building a SKILL.md that encodes their domain expertise. This is not a competency assessment, not a process documentation exercise, and not a performance review. It is a collaborative effort to capture the professional judgement that makes their work valuable: the knowledge that their junior colleagues are trying to absorb and that no manual currently documents.
Describe the output. The output will be a deployable agent they will have the opportunity to test and refine. This matters because it shifts the expert's motivation from "help the interviewer understand my job" to "help build something I want to work." Experts who understand the output invest more effort in specificity because they can see how vague answers will produce a vague agent.
Set the process expectation. The interview is the starting point of a process, not the end of one. The first-draft SKILL.md will be tested against scenarios, and the expert will review the results and refine the instructions. This framing reduces the pressure to get everything perfect in a single conversation; which paradoxically makes the expert more willing to share uncertain or provisional knowledge rather than limiting themselves to what they are certain about.
The briefing matters because it determines which version of the expert's knowledge you access. Without it, the expert defaults to performance mode: presenting their expertise as they would in a job interview, an industry panel, or a training presentation. Performance mode produces the official narrative: polished, accurate at the headline level, and missing the operational specificity that a SKILL.md needs. With the briefing, the expert enters collaborative mode: helping you build something they want to work. Collaborative mode produces operational knowledge: the specific examples, the failure patterns, the heuristics that exist below the official narrative.
The goal during the interview is not to transcribe every word. It is to capture the specific examples, the specific failure modes, and the specific heuristics that appear in answers to the five questions. The distinction between specific and generic material is the most important note-taking skill in the extraction process.
What You Hear
Type
What To Do
"We always prioritise risk management"
Generic
Follow up: "Can you give me a recent example where risk management changed what you actually did?"
"When the earnings release date is within ninety days, we run worst-case sensitivity rather than base case"
Specific
Capture verbatim: this is a candidate SKILL.md Principle
"Experience matters a lot in this work"
Generic
Follow up: "What does experience let you see that someone without it misses? Can you give me an example?"
"I look for the gap between what the management says about working capital and what the cashflow statement shows"
Specific
Capture verbatim: this is decision-making logic
"It depends on the situation"
Generic (but promising)
Follow up: "Walk me through two situations where it went differently. What was different about the context?"
The pattern is consistent. Generic statements are not failures: they are signals that a follow-up question is needed. Underneath every generic statement is a specific one. "We always prioritise risk management" becomes "when earnings are within ninety days, we run worst-case sensitivity." The follow-up question is what makes the transition happen.
Record the interview if the expert agrees and if you have a practical transcription mechanism available. But do not rely on the recording as a substitute for real-time note-taking. The act of capturing specific statements in the moment forces you to listen for specificity, which in turn shapes the follow-up questions you ask. Interviewers who rely on recordings tend to ask fewer follow-up questions because they assume they can extract the detail later: and the detail is often not there because it was never surfaced.
The five questions are a framework, not a script. Allow the conversation to go where the expert's experience takes it. If Question 2 produces a long and detailed answer that covers material relevant to Questions 3 and 4, follow it rather than interrupting to keep to the sequence. The sequence is designed to ensure coverage, not to impose a rigid structure.
The coverage check comes at the end of the interview, not during it. After the expert has finished responding to the fifth question (or after the natural conclusion of the conversation, whichever comes first) review your notes against the five-question framework and identify any areas that were not covered. If Question 3 was only lightly touched because Questions 2 and 4 absorbed the material, that is fine. If Question 5 was never addressed and the automation boundaries are missing from your notes, ask it before the conversation ends.
There is one conversational discipline that matters more than maintaining the question sequence: redirecting from general to specific. When the expert moves from a specific account to a general principle, "and that is why we always check the covenant documentation before making a credit decision", acknowledge the principle and then redirect back to the specific: "Can you give me an example where checking the covenant documentation changed your assessment? What would have happened if you had not checked it?"
The redirect is gentle and consistent. You are not challenging the expert's general statements. You are using them as springboards to reach the specific knowledge that sits underneath.
After the interview, before you begin writing the SKILL.md, write a two-paragraph summary of what you learned. Do this immediately: within thirty minutes of the interview ending, while the specific details are still fresh. The quality of this synthesis degrades rapidly with time. A summary written the next day is noticeably less precise than one written within the hour.
The first paragraph captures the most important decision-making logic the interview surfaced. What is the core analytical process the expert applies? What are the key signals they look for? What is the sequence in which they evaluate information?
The second paragraph captures the most important escalation condition the interview identified. What are the situations where human judgement is genuinely irreplaceable? Where are the boundaries of what the agent should handle autonomously?
Credit analyst north star summary example:
The most important decision-making logic this interview surfaced is the analyst's practice of reading cashflow statements before balance sheets and checking revenue quality through working capital cycle analysis rather than trusting headline figures. When receivables days increase while revenue is flat, the analyst treats revenue as weakening regardless of what the income statement shows. Contracts underlying pre-sale figures are read for conditionality rather than taken at face value. When the management narrative and the financial data diverge, the analyst trusts the data.
The critical escalation condition is a three-part boundary. Credit decisions above £25 million go to the senior committee regardless of analysis quality. Assessments involving borrowers connected to board members or senior executives are routed to an independent reviewer. Any fact pattern the analyst has not previously encountered: a novel industry structure, an unfamiliar regulatory regime: is flagged and referred to a specialist rather than assessed using a framework that may not apply.
This summary is your north star for the first draft. If the SKILL.md you write does not clearly encode the substance of both paragraphs, something has been lost in the translation from interview to instructions. Every revision of the SKILL.md should be checked against it.
Use these prompts in Anthropic Cowork or your preferred AI assistant to practise the interview execution skills.
What you're learning: Briefing is a skill that adapts to context. A compliance officer has different concerns about the interview process than an architect or a clinician. Practising the briefing language for a specific role builds the adaptive communication skill you need to conduct effective extraction interviews across professional domains.
What you're learning: The ability to transform generic statements into specific SKILL.md Principles is the core note-taking skill in knowledge extraction. Every generic statement contains a specific one: the follow-up question is what surfaces it. This exercise trains you to recognise generic language instantly and respond with the question that produces actionable material.
What you're learning: The north star summary is the bridge between the interview and the SKILL.md. Writing it well requires distinguishing the most important extraction material from the rest: a judgement that improves with practice. Assessing your own summary's specificity also trains you to evaluate whether your interview produced enough actionable material or whether a follow-up session is needed.
The Method A interview is a structured conversation, not a form: and its success depends on three operational elements: a briefing protocol that establishes collaborative mode, a note-taking approach that captures specific rather than generic knowledge, and a north star summary written immediately after that anchors the first SKILL.md draft.
📋Quick Reference
Access condensed key takeaways and quick reference notes for efficient review.
Free forever. No credit card required.