Imported from arashbehmand/phoenix-skills (skills/preparing-for-interviews/SKILL.md). Install upstream with npx skills add arashbehmand/phoenix-skills --skill preparing-for-interviews. Copyright stays with the author.

Preparing for interviews

Produce a preparation guide and write it to applications/<slug>/interview-prep.md.

The point is a guide the candidate could not have written for themselves in an hour: the questions they will actually be asked, tied to the requirements those questions test, and the stories they should have ready. Generic advice — "research the company", "be confident" — is worse than nothing, because it uses up the time they had.

Read first

Ask which round it is and who is in it if the workspace does not say. The right guide for a recruiter screen and for a final panel are not the same document.

The rule that keeps this honest

Anti-Hallucination: Do not invent company values, news, or interviewer details not present in the Input. If data is missing (e.g., interviewer name), state that you cannot provide specific advice for it, rather than guessing.

This is the one that gets broken, because a guide with a confident "Interviewer X will care about system design" reads better than one that says the interviewer is unknown. It is also the one whose failure costs most: a candidate who walks in having rehearsed invented company values will use them, and it will land badly.

In practice:

• No company research? Section 1's company entry says so. Do not reconstruct values from the careers page's marketing copy.
• No interviewer information? Say it is not available and suggest asking the recruiter for names and focus areas — a normal request that usually gets answered — rather than guessing what someone cares about.
• Anything inferred rather than known gets ⚠️ and a note saying what it rests on. A question built on "two leads appear to have left, per LinkedIn" is speculative; label it, because the candidate must not raise it as established fact.

Better to hand back a guide with a visible hole in it than a complete-looking one with an invented patch.

Do not write down what you left out. A guide saying "no private details (X, Y, Z) were included" has just written X, Y and Z into a third file, next to the ones that get sent. "Private material excluded" carries the same assurance and names nothing. Usually nobody asked.

Output

Write applications/<slug>/interview-prep.md in this structure.

# 🎯 Personalized Interview Preparation Guide

*<Company> — <Role> · <which round, format, date>*
*Generated <date> · inputs: <files actually read>*

## 0. Executive Summary & Top Priorities

*   **Top Priority 1:** (specific, e.g. rehearse the X story because it matches
    requirement Y)
*   **Top Priority 2:**
*   **Top Priority 3:**
*   **Estimated Prep Time:** [e.g. "6-8 hours over the next 3 days"]

## 1. Essential Knowledge & Strategic Checklist

*   **Company Deep Dive:** (mission, values, products, recent news, culture, competitors)
*   **Job Role Mastery:** (responsibilities, required vs desired skills, team, impact)
*   **Résumé Alignment & Storytelling:** (experiences mapped to requirements, STAR stories)
*   **Interviewer Insights & Research:** (their role, focus, questions to ask them)
*   **Logistics & Etiquette:** (format, timing, platform, follow-up plan)
*   **Mindset & Confidence:** (specific to this candidate, not general encouragement)

## 2. Top 20 Tailored Interview Questions

*Exactly 20, each with a one-line rationale naming the requirement it tests.*

1. "..." (Rationale: tests the 'X' requirement in the job description.)
...
20. "..."

*(Behavioural: 6 | Technical/Role-Specific: 5 | Situational: 4 | Company/Role Fit: 3 |
Career Vision: 2)*

### Honourable Mentions

*5-10 more, including curveballs. ⚠️ flags anything built on speculative data.*

## 3. Strategic Insights & Beyond the Ordinary

*   **Anticipated Format & Unspoken Expectations:**
*   **Interviewer-Specific Angles:**
*   **Common Pitfalls to Avoid:**

## 4. Crafting Your Narrative: Key Talking Points

*3-5 story frameworks. Each maps to a named job requirement AND a company value.*

*   **Talking Point 1 — <theme>**
    *   **Theme:**
    *   **Connects to:** JD requirement ("...") & Company Value ("...")
    *   **Story Snippet:** "..." (written out, in the candidate's voice)

## 5. Extended Preparation Toolkit

*   **Targeted Mock Interviews:**
*   **"Elevator Pitch" Refinement:**
*   **Portfolio/Project Deep Dive:**
*   **Industry-Specific Research:**

---

### AI Confidence, Disclaimer & Feedback Loop

*   **Guide Confidence Score:** [High/Medium/Low] — and what would raise it.
*   **Disclaimer:** Generated from the inputs listed above. No real-time data, no human
    intuition. A supplement to your own research, not a substitute.
*   **Next Steps:** Offer to go deeper on any section, or to regenerate when more is known.

The parts that carry the weight

The question mix is enforced: 6 behavioural, 5 technical, 4 situational, 3 fit, 2 career vision. Twenty questions, and the counts are stated at the end of the section. Without the quota a generated list drifts into whichever category is easiest, which is usually behavioural — and then the candidate is unprepared for the four situational questions that decide the round.

Every question carries a rationale naming a real requirement. "Rationale: tests the 'Stakeholder Management' skill listed in the JD." This is what makes the list preparation rather than trivia — the candidate can see which requirement each question is probing and answer the requirement, not the question.

Include the questions they will not enjoy. The gap from fit.md. The five years at one employer with no promotion. The thing honest-context.md says they dread. Being asked for the first time in the room is far worse than reading it here. If preferences.md asks for it harsh, honour that. references/hard-questions.md is the recurring set — why you left, the gap, the short stints, salary, right to work — and how to build a true, short, non-defensive answer to each from what the honest context already says.

Talking points map to a requirement and a company value, and the story snippet is written out rather than described. A candidate cannot rehearse "prepare a story about cross-functional work"; they can rehearse three sentences. Write them in the candidate's voice, from their real experience, with their real numbers. If a value cannot be sourced, map to the requirement alone and say the value is unknown.

Estimated Prep Time is a real estimate — hours and a window. It tells the candidate whether to start tonight.

Guide Confidence Score reflects the inputs, not the writing. Say what would raise it: usually the interviewer's name, or company research.

Tone

Blunt, compact, honest. No sugar-coating. If a critical skill is missing, flag it. No generic advice — every item should be something the candidate could not have written themselves, specific to this role, this company and this résumé.

Assume they are in a hurry and reading this the night before.

If the environment has more tools

Nothing here requires any of them, and the anti-hallucination rule above outranks all of them — a tool that returns nothing means the data is missing, not that you should guess.

• Web search or fetch — the honest way to fill the company section when research/<company>.md does not exist. Prefer running researching-companies properly and writing the file, so the next skill benefits too.
• A LinkedIn MCP server, if the user already runs one — an interviewer's public role and background is the single input that moves the confidence score most. Read only what the person has published, do not infer from activity, and say where it came from. The simpler route is usually asking the recruiter for names and focus areas.
• Calendar access — only to confirm the time and format you were told about.

After writing

Put the interview in the ## Timeline in notes.md as a dated entry, with the time if there is one — that is what makes it show up in a review of the week — and add a dated line for the guide itself. If anything was deliberately left blank, an unknown interviewer or missing research, put it under ## Next so it is obvious what to fill in when the recruiter replies.

Offer to run a mock interview afterwards. Reading twenty questions is not the same as having answered them out loud, and the candidate can only find the sentences that do not work by saying them.

Next

• Post-interview follow-up, or a recruiter reply → drafting-outreach-replies
• Company research was missing → researching-companies
• The round changed the picture → revisit assessing-job-fit