Skill: expert-interview-guide

Preconditions

• Input mode chosen: document_attached (proposal / scope doc supplied) or manual (Q&A captures the inputs below).
• docx skill available if generating .docx (or markdown fallback if not).
• For document_attached mode: PDF / DOCX / PPTX readable.
• Operator can answer or has documented: engagement type, interviewee type, sector, top 2-3 priority topic areas.

When to use

Operator is preparing an interview — primary research, customer due diligence, competitor benchmark, expert call, stakeholder mapping. The skill is generic across sectors and engagement types.

NOT for:

• Internal meeting prep → use prep-my-week or meeting-close.
• Journalistic interview → out of scope.
• Survey / questionnaire at scale → not 1:1, different shape.

Configuration (edit before first use)

default_project_template: ""             # e.g. "DD", "MarketStudy", "GrowthStrategy"
default_language: "en"
internal_disclaimer: "INTERNAL USE ONLY — DO NOT DISTRIBUTE EXTERNALLY"
default_duration_min: 60
default_priority_split:
  must_have: 12
  important: 6
  optional: 4

Input contract

input_mode: document_attached | manual

# if input_mode=document_attached
proposal_path: str | None      # PDF / DOCX / PPTX with scope or proposal
                               # extract: project, target, scope topics, hypotheses

# if input_mode=manual
engagement_type: commercial_dd | vendor_dd | market_study | growth_strategy | operational_dd | other
interviewee_type: customer | competitor | industry_expert | channel_partner | supplier
duration_min: 30 | 45 | 60 | 90
project_name: str
target_company: str | None     # [Company] placeholder if absent
sector: str
geography: str | None

topic_areas: list[str]         # e.g. ["market_sizing", "competitive", "pricing", "tech"]
priority_areas: list[str]      # 2-3 highest priority — drive MUST-HAVE tagging
hypotheses: list[str] | None   # specific hypotheses to test

output_dir: str | auto
language: str                  # default: configured
generate_prep_brief: bool      # default: true

Output contract

guide_path: str                # clean .docx (or .md fallback)
prep_brief_path: str | None    # internal-only prep brief, null if disabled
question_count: int
must_have_count: int
important_count: int
optional_count: int

Sequence

Step 1 — Resolve inputs

• document_attached: read the proposal, extract project / target / topics / hypotheses. Confirm extracted fields with the operator before generating.
• manual: Q&A walks the operator through the input contract.

Step 2 — Calibrate question count by duration

Step 3 — Generate the clean guide

Structure:

1. Header — interview metadata (date placeholder, project, interviewee role, duration). No internal disclaimer.
2. Warm-up (2-3 min) — role, tenure, scope. Non-threatening, gets the interviewee talking.
3. Topic sections — one per topic_area. Within each: questions in priority order.
4. Closing (3-5 min) — "what else?", referrals, follow-up permission.

Each question:

• Carries a priority tag: [MUST-HAVE] / [IMPORTANT] / [OPTIONAL].
• Carries [Quantify] on every metric question so the interviewer knows to push for numbers.
• Open-ended by default (avoid yes/no unless the goal is confirmation).
• One concept per question (no compound).

Step 4 — Generate the prep brief (internal-only)

Mirror the guide's structure, but for each topic section add:

• Hypothesis under test (from hypotheses input or inferred from priority_areas)
• Benchmark data — known industry numbers / prior interview findings, with source
• Anchoring — what the interviewer should already believe, and what would change it
• Contradiction watch — claims to expect from this interviewee that contradict known data; how to probe without antagonizing
• Internal disclaimer at top — INTERNAL USE ONLY — DO NOT DISTRIBUTE

If generate_prep_brief: false, skip Step 4.

Step 5 — Output

• Save guide to output_dir/<project>-<interviewee>-guide-YYYY-MM-DD.docx
• Save prep brief to output_dir/<project>-<interviewee>-prep-internal-YYYY-MM-DD.docx
• Return file paths + question count breakdown

Anti-patterns

• Confusing the two outputs — never paste prep-brief content into the clean guide; the internal disclaimer is the only safeguard against accidental external distribution.
• Closed questions on quantifiable topics — "do you use X?" wastes a turn vs "how much X do you use, and how has that changed?".
• One section per priority instead of per topic — fragments the interview flow. Topics are the unit; priorities are tags within topics.
• Forgetting the [Quantify] marker on metric questions — the interviewer drifts to qualitative comfort if not prompted.
• Skipping the prep brief for short interviews — even 30-min interviews benefit from explicit hypotheses + benchmarks. The brief is short for short interviews, not absent.