beam-ai-team/interview-coach

Interview Coach

Honest feedback to help you become a better interviewer.

Safety Contract

Before fetching interview transcripts, reading job descriptions, saving feedback, or pushing reviews to Workable, show the candidate, role, transcript source, JD source, destination/write target, and sensitive hiring data included. Require explicit user approval in the current turn before fetching from external systems, writing files, or updating ATS records. In-chat coaching based on user-pasted text does not require approval beyond confirming the data source.

When to use: After conducting an interview, get your transcript analyzed for actionable feedback

Prerequisites:

• fathom-fetch-meetings or Amie — Get interview transcript
• workable-master — Pull job description + push review to Workable

Output: Expert panel assessment, barrel score, Workable review

---

Purpose

Help hiring managers become better interviewers through structured, evidence-based feedback. Instead of relying on gut feelings or informal peer reviews, this skill analyzes actual interview transcripts against the job description to identify specific patterns — what's working, what's missing, and how to improve. The goal is consistent, fair interviews that assess candidates accurately while creating a positive candidate experience.

---

Workflow

Step 1: Get the Transcript

Option A — By candidate name (recommended):

User: "review my interview with Sarah"
→ AI searches Fathom for meetings with "Sarah" in title
→ If multiple matches, shows list to pick from
→ Fetches transcript automatically

Option B — Manual fetch:

User: "fetch meetings from fathom"
→ Lists recent meetings
→ User selects one
→ Then says "review my interview"

Fathom search by name:

# Fetch recent meetings, filter by title containing candidate name
curl -s --request GET \
  --url 'https://api.fathom.ai/external/v1/meetings?include_summary=true&created_after={30_DAYS_AGO}' \
  --header 'X-Api-Key: {FATHOM_API_KEY}' \
  | jq '.meetings[] | select(.title | ascii_downcase | contains("{candidate_name}"))'

Link to prerequisite skill: fathom-fetch-meetings fetches meetings filtered by attendee domain, returns transcripts.

Step 2: Get the Job Description

Once transcript is available, fetch the JD from Workable:

Ask: "What role was this interview for?"

Then run:

python3 03-skills/workable-master/scripts/workable_client.py --fetch-jd "[role title]"

If multiple matches, show options and ask for shortcode:

python3 03-skills/workable-master/scripts/workable_client.py --fetch-jd-code [shortcode]

Link to skill: workable-master handles all Workable operations.

Step 3: Interview Stage (REQUIRED)

Ask: "What type of interview was this?"

| Type | When to Use | |------|-------------| | hr | HR/Cultural round — soft skills, culture fit, resume validation | | technical | Technical/Case Study round — in-depth skill assessment, problem solving | | final | Final round — career aspirations, team fit, leadership potential |

Triggers can specify type directly:

"hr round review with Sarah"
"technical round review with John"
"final round review with Ahmed"

---

Evaluation Frameworks

Each interview type has its own evaluation criteria. The framework ensures you're assessing what matters for that stage.

---

🟢 HR Round Framework

Focus: Cultural fit, soft skills, resume-personality alignment

| Dimension | Weight | What to Assess | |-----------|--------|----------------| | Cultural Fit | 25% | Values alignment, work style, company culture match | | Resume-Personality Match | 20% | Does their presence match their paper claims? Authenticity check | | Soft Skills | 20% | Communication clarity, adaptability, emotional intelligence | | Motivation & Fit | 15% | Why this company? Career trajectory alignment | | Red Flag Detection | 10% | Inconsistencies, concerning patterns, gaps explained | | Candidate Experience | 10% | Did you create a positive first impression? |

Key Questions for HR Round:

• "What kind of work environment brings out your best?"
• "Tell me about a time you disagreed with a colleague"
• "What attracted you to this role specifically?"
• "Walk me through your career journey"
• "How do you handle feedback?"

Red Flags to Probe:

• Badmouthing previous employers
• Unexplained gaps or frequent job changes
• Vague answers about leaving previous roles
• Mismatches between resume claims and conversation

---

🔵 Technical/Case Study Round Framework

Focus: Deep skill assessment, problem-solving ability, technical communication

| Dimension | Weight | What to Assess | |-----------|--------|----------------| | Technical Depth | 25% | Did you probe actual skills at required depth? | | Problem-Solving Approach | 25% | Did you assess methodology, not just answers? | | Skill Validation | 20% | Did questions match JD requirements? Evidence-based assessment | | Technical Communication | 15% | Can they explain complex concepts clearly? | | Real-World Application | 10% | Did you test practical knowledge, not just theory? | | Assessment Fairness | 5% | Equal opportunity to demonstrate, hints when stuck |

Key Questions for Technical Round:

• "Walk me through how you'd approach [specific problem]"
• "What's the trade-off between X and Y approach?"
• "Tell me about a technical challenge you solved recently"
• "How would you debug [scenario]?"
• "Explain [concept] to a non-technical stakeholder"

Case Study Best Practices:

• Give candidate time to think (don't rush)
• Probe reasoning, not just final answer
• Ask "what would you do differently?"
• Test both breadth and depth

---

🟣 Final Round Framework

Focus: Career aspirations, team fit, leadership potential, mutual evaluation

| Dimension | Weight | What to Assess | |-----------|--------|----------------| | Career Aspirations Alignment | 25% | Does their growth path match what you can offer? | | Team Fit | 25% | How will they mesh with existing team dynamics? | | Leadership Potential | 20% | Strategic thinking, influence, decision-making maturity | | Long-term Commitment Signals | 15% | Are they building a career or just taking a job? | | Mutual Fit Exploration | 10% | Did you honestly discuss role realities? Two-way evaluation | | Closing Effectiveness | 5% | Did you sell the opportunity and address concerns? |

Key Questions for Final Round:

• "Where do you see yourself in 3 years?"
• "What would make you turn down this offer?"
• "How do you like to be managed?"
• "Tell me about a time you influenced without authority"
• "What questions do you have for me about the team?"

Final Round Focus:

• This is about mutual fit, not skills (already validated)
• Be honest about challenges and expectations
• Assess culture add, not just culture fit
• Leave time for their deep questions

---

General Evaluation Framework

For interviews not fitting the above categories, use the general 6-dimension framework:

Analyze across 6 dimensions, then calculate overall score out of 5:

1. Role Alignment

Do questions match the JD?

• Questions probe JD requirements
• Technical depth appropriate for role
• Scenarios match actual job challenges

2. Question Quality

Open-ended, behavioral, good follow-ups?

• Open vs closed questions ratio
• Behavioral questions (STAR-friendly)
• Follow-up probes ("Tell me more", "What happened next")
• Flag: Leading questions ("You're good with deadlines, right?")

3. Bias Signals

Fair and consistent?

Red flags:

• Assumptions based on background
• Interrupting candidate frequently
• Affinity bias ("Oh, you went to my school too!")
• Questions about protected characteristics

4. Structure & Flow

Clear progression?

• Opening: Set expectations, build rapport
• Middle: Logical topic flow
• Closing: Time for candidate questions, next steps explained
• Time management: Not rushed at end

5. Talk Ratio

Candidate should talk 70-80%

• Count interviewer vs candidate speaking time
• Flag if interviewer talks >40%
• Note long interviewer monologues

6. Candidate Experience

Would they recommend interviewing here?

• Warmth and rapport
• Active listening
• Selling the role appropriately
• Clear communication about process

---

Scoring

Overall Score: X/5

| Score | Meaning | |-------|---------| | 5 | Excellent — nothing significant to improve | | 4 | Good — minor refinements will make you great | | 3 | Adequate — clear areas to work on | | 2 | Needs improvement — several patterns to address | | 1 | Significant concerns — fundamental issues to fix |

Calculate: Average across 6 dimensions, round to nearest 0.5

---

Output Format

Uses the standard review template from workable-master SKILL.md — plain-text format with:

1. Expert Panel Assessment — 3 experts, bullet points (Strong/Concern), score per expert
2. Barrel Scoring — 5 dimensions, classification
3. Candidate Scorecard — 7 dimensions scored /10, overall average
4. Interview Quality — short assessment of whether the questions were good, time was managed well, and gaps were probed
5. Recommendation — clear verdict with next steps

The Interview Quality section is unique to this skill (not in candidate-compare). It evaluates the interviewer's performance:

━━━ INTERVIEW QUALITY ━━━

• Questions matched JD: [Yes/Partially/No] — [brief note]
• Failure probing: [Strong/Adequate/Weak] — [brief note]
• Case question quality: [Strong/Adequate/Weak] — [brief note]
• Time management: [Good/Okay/Poor] — [brief note]

This gives the interviewer quick feedback on their own performance alongside the candidate assessment.

---

Best Practices Reference

Strong Questions

Behavioral (STAR):

• "Tell me about a time when..."
• "Describe a situation where you had to..."

Situational:

• "How would you handle..."
• "What would you do if..."

Follow-ups:

• "What was the result?"
• "What would you do differently?"

Avoid These

| Avoid | Instead | |-------|---------| | "Are you a team player?" | "Tell me about a time you collaborated..." | | "What's your greatest weakness?" | "What skill are you actively developing?" | | "Can you work under pressure?" | "Describe a high-pressure situation you handled" |

Ideal Structure

[0-5 min]   Welcome & rapport, explain format
[5-35 min]  Core questions with follow-ups
[35-45 min] Candidate questions
[45-50 min] Close — next steps, timeline, thank them

---

Tone

Be a coach, not a critic:

• "Here's an opportunity..." not "You did this wrong"
• "Consider trying..." not "You should have..."
• Always quote the transcript
• Provide concrete alternatives

---

Triggers

| User Says | Action | |-----------|--------| | "review my interview with [name]" | Search Fathom by name, fetch transcript, ask for interview type | | "hr round review with [name]" | Fetch transcript, analyze with HR framework | | "technical round review with [name]" | Fetch transcript, analyze with Technical framework | | "final round review with [name]" | Fetch transcript, analyze with Final framework | | "assess my interview with [name]" | Same as review | | "how's the candidate [name]" | Same as review | | "review my interview" | Ask for candidate name and interview type | | "assess my interview" | Same | | "interview feedback" | Same | | "interview coach" | Same | | "hr round review" | Ask for candidate, use HR framework | | "technical round review" | Ask for candidate, use Technical framework | | "final round review" | Ask for candidate, use Final framework | | Pastes transcript | Ask for role and interview type, then analyze |

---

Post-Evaluation: Expert Assessment + Push to Workable

After generating the interview coaching feedback, always produce the candidate review using the standard template from workable-master:

Automatic Steps:

1. Select 3 experts based on the role (see workable-master SKILL.md for role→expert mapping)
2. Run expert assessments — each expert evaluates the candidate from their lens
3. Run barrel scoring — 5 dimensions, weighted total, classification
4. Compile strengths, gaps, and follow-up questions
5. Determine grade — map expert consensus + barrel score to Yes/Maybe/No

Then offer to push:

💡 Push this review to Sarah's Workable profile? (Grade: Yes ✅)

When user confirms:

python3 03-skills/workable-master/scripts/workable_client.py \
  --candidate "{candidate_name}" \
  --review --grade {0|1|2} \
  --assessment "{full_review_text}"

If user declines: Still save the review locally but don't push.

---

Integration

Connected skills: | Skill | Purpose | |-------|---------| | fathom-fetch-meetings | Get interview transcript (Fathom) | | Amie notes/transcripts | Get interview transcript (Amie) | | workable-master | Fetch JD + push review to Workable |

Full Workflow (end-to-end):

Manager: "review my interview with Sarah"
→ AI searches Fathom for "Sarah" in meeting titles
→ AI: "What role was this interview for?"
Manager: "Product Manager"
→ AI fetches JD from Workable (workable-master)
→ AI selects 3 expert panel based on role
→ Expert assessments + barrel scoring
→ AI: "Push this review to Sarah's Workable profile? (Grade: Yes ✅)"
Manager: "yes"
→ Review posted to Workable with grade

Alternative (step-by-step):

Manager: "fetch meetings from fathom"
→ Lists recent meetings
Manager: picks meeting
Manager: "review my interview"
→ continues as above

No manual transcript or JD pasting required — everything pulls from your systems.

---

Version: 2.0 | Updated: 2026-02-16

Changelog:

• v1.6: Workable push integration — after evaluation, offers to push assessment to candidate's Workable profile via workable-push-assessment
• v1.5: Stage-specific evaluation — added HR Round, Technical Round, and Final Round frameworks with tailored dimensions and weights
• v1.4: Added candidate name search — "review my interview with [name]" fetches transcript automatically
• v1.3: Integrated with workable-fetch-jd for automatic JD fetching
• v1.2: Added triggers "assess my interview", "how's the candidate"
• v1.1: Integrated with fathom-fetch-meetings, simplified to 5-point scale, removed letter grades
• v1.0: Initial creation