# Prepare for Interviews

Source: https://career-ops.org/docs/introduction/guides/interview-modes (canonical HTML, identical content)

> Use the interview modes — plan, practice, debrief — to prepare for a specific interview round, rehearse it with an AI interviewer, and turn every real interview into intelligence for the next one.



The interview modes are career-ops's interview preparation system, shipped in v1.16.0: `interview/plan` builds a time-blocked prep plan for a specific round, `interview/practice` runs a realistic mock interview one question at a time, and `interview/debrief` captures what actually happened so the next round starts smarter. Together they form a loop — plan, practice, interview, debrief, repeat — that compounds across a hiring process.

They complement the older [`interview-prep` mode](https://career-ops.org/docs/reference/modes/interview-prep.md), which builds your durable foundation (story bank, question bank). The three focused modes are about a *specific upcoming round*.

> **What you need first**
A workspace with your `cv.md` and `config/profile.yml` in place (the [Quick Start](https://career-ops.org/docs.md) sets both up). The modes read them to ground every plan and every piece of feedback in your real experience — not generic interview advice.

## The loop at a glance [#the-loop-at-a-glance]

| Mode                 | When                            | What you get                                                   |
| -------------------- | ------------------------------- | -------------------------------------------------------------- |
| `interview/plan`     | The moment a round is scheduled | A time-blocked prep plan sized to the hours you actually have  |
| `interview/practice` | Before the round                | A realistic mock interview with structured feedback per answer |
| `interview/debrief`  | Right after the real interview  | An honest per-question assessment + an updated question bank   |

## Plan: turn a date into a schedule [#plan-turn-a-date-into-a-schedule]

Give `interview/plan` the job description and the interview date and time, plus the round type and interviewer if you know them. It reads your CV and profile, runs an honest fit assessment — strengths to anchor on, gaps likely to be tested — and divides the hours you have left into concrete prep blocks.

```txt title="Example"
/career-ops interview/plan

→ Inputs: JD (pasted), interview Thursday 14:00, HM screen with Ana (Eng Manager)
→ Fit assessment: 4 strengths to anchor · 2 gaps ranked by test likelihood
→ Plan (9 hours available):
   Block 1 — Lock your narrative (~1.5h): timeline, "why us", strongest proof point
   Block 2 — Gap: event-driven architectures (~3h)
   Block 3 — Behavioral stories, STAR+R (~2h)
   Block 4 — Mock run + comp answer (~1.5h)
```

The plan calibrates to the round: a recruiter screen gets logistics and narrative, not deep internals; a practitioner deep-dive gets depth in the core skill from the JD. If a question bank from a previous round exists, anything you struggled with gets a dedicated block — real performance data outranks inferred risk.

## Practice: rehearse against a realistic interviewer [#practice-rehearse-against-a-realistic-interviewer]

`interview/practice` plays the interviewer — by name and role if you provide them — and asks one question at a time, staying in character, with natural follow-ups when a real interviewer would push deeper. After each answer you get structured feedback: what landed, what to tighten, and a stronger version of your answer grounded in your actual CV and story bank.

It also tracks the session the way an interviewer would: if you reach for the same story twice, it flags that your example set reads thin; if your answer closes on the wrong domain for the role, it tells you where to land it instead.

Practice is honest about its own limits: if you haven't built a question bank or role-specific prep notes yet, it says so and offers to run `interview-prep` or `interview/plan` first, rather than silently running a generic session.

## Debrief: never lose an interview's intelligence [#debrief-never-lose-an-interviews-intelligence]

Run `interview/debrief` right after a real interview, while memory is fresh. It walks you through every question you can recall, records how the interviewer reacted, and produces an honest per-question assessment. Then it does the compounding work:

* **Updates the question bank** — each question marked ✅ / 🟡 / 🔴 based on real performance, not gut feeling.
* **Closes the gaps** — for every 🔴, it explains the correct answer with a worked example and connects it to a story you already have.
* **Extracts new stories** — if you told something in the room that isn't in your story bank yet, it builds it into a STAR+R story on the spot.
* **Predicts the next round** — if you know the format, it forecasts likely questions from the interviewer's role and what this round already covered.

## Why this loop works [#why-this-loop-works]

Interview prep usually resets to zero with every round. The interview modes make it cumulative: what a real interviewer actually asked becomes the highest-priority prep signal for the next round, stored in plain Markdown files (`interview-prep/question-bank.md`, `story-bank.md`) that you own, on your machine, like everything else in career-ops.