meeting-qualifier

---
name: meeting-qualifier
description: Score a prospect from public evidence before accepting the meeting, then bring only the questions that close the gaps. Use before booking, preparing or writing up a sales meeting.
---

# Know before the meeting whether it can close

A scorecard to accept, shorten or decline, ten gap-closing questions in the order that earns budget talk, one page afterwards.

## Claude does

Inputs: {SELLER}, what the user sells; {PROSPECT}, name, company, role, sector, how the meeting arose, first or repeat; {EVIDENCE}, emails or earlier notes; {STAGE}, pre-meeting or post-call; {ANSWERS}, post-call notes.

1. Read the prospect's website and {EVIDENCE}; write {PROSPECT_BRIEF}: what they do, their sector's words, the likely trigger. "Unknown" beats a guess.
2. Score Budget, Authority, Need, Timeline 0-25 each from evidence only, every signal tagged High, Medium, Low or Inferred with its quoted source. MEDDIC completeness: elements at Medium or better out of six, as a percentage; 75 or more is grade A. Record {SCORECARD}.
3. {ACTION}: grade A, accept and prepare the next step; below A, discovery only, the gaps are the agenda; no pain and no trigger anywhere, a short call, not an hour.
4. {GAPS}: elements lacking buyer evidence, most valuable first. {KEY_QUESTION} closes the biggest; {TRAP} is the likeliest mistake on this call.
5. Post-call: re-score from {ANSWERS} (buyer's words High, impressions Low). {NEXT_STEP}: grade A, proposal or the natural next meeting; below A, a meeting on the remaining gaps, no proposal; Need still Low or Inferred, none. {BIGGEST_RISK}: one sentence. Failed: {EXIT_POINTER}, somewhere genuinely useful to send them; else "none".
6. Ask once for anything missing; never invent it.

## Then Codex does

Codex formats the scorecard, drafts the ten questions with four planning fields from the bank in CODEX.md in sector language, the cheat sheet and, post-call, the note and exit letter: fixed shapes from decisions already taken.

```
codex exec --sandbox danger-full-access --skip-git-repo-check -C "<working folder>" - < CODEX.md
```

First Claude writes a filled CODEX.md into the working folder, every placeholder named above replaced.

## Claude checks

1. Ten questions in the fixed order; budget and authority never before pain and impact.
2. Each has Purpose, Expected Response, Follow-Up, Listen For; big ones open "Can you walk me through" or "Can you help me understand"; no banned opener; no wording that implies its answer.
3. Files match {SCORECARD} and {NEXT_STEP}; nothing Medium or above without a quoted source; no fact beyond the inputs; graceful-exit.md only when {EXIT_POINTER} is not "none".
4. Any failure: rerun Codex naming the check, check again; show nothing until all pass.

## Rules

- Ask, then listen: questions, not a script to read at them.
- Pain and impact before budget or authority; never expect full BANT on call one.
- Only questions you genuinely want answered; nothing leading, no pressure.
- Plan ten, ask at most fourteen, under a tenth on product; no demo, no interrogation.
- Banned openers: "Tell me about your company", "How's business?", "What are your goals?", "What keeps you up at night?", "Are you the decision-maker?"
- Gaps only; senior buyers have already sat through two or three discovery calls.
- An untested coach is not a Champion; nobody is economic buyer until someone names who signs.
- Nothing goes green without buyer evidence; Inferred is a gap; nobody buys without pain.
- Repeats open with "What has changed since we last talked?"; always ask who else is involved and who signs.
- Failed prospects get a graceful exit: a useful pointer elsewhere, door open.

## If Codex is not installed
Claude does the "Then Codex does" section itself, following CODEX.md, and says so.

## Built from

- AI Sales Team for Claude Code: https://github.com/zubair-trabzada/ai-sales-team-claude/blob/main/skills/sales-prep/SKILL.md (1,093 stars, api.github.com)
- Anthropic knowledge-work-plugins, call-prep: https://github.com/anthropics/knowledge-work-plugins/blob/main/sales/skills/call-prep/SKILL.md (23,876 stars, api.github.com)
- GitLab Handbook, MEDDPPICC: https://handbook.gitlab.com/handbook/sales/meddppicc/ (a public company's live playbook)
- Gong, 26 MEDDIC Questions: https://www.gong.io/files/gong-guide-meddic-questions.pdf (plus Gong Labs, 519,000+ calls)
- 30MPC, Brian LaManna: https://www.30mpc.com/newsletter/the-8-discovery-questions-that-got-me-to-presidents-club (six-time President's Club)
- HubSpot, How I Use BANT: https://blog.hubspot.com/sales/bant (most-cited BANT reference)

# Know before the meeting whether it can close - Codex task

Draft from the inputs only. No research, no added facts, no changed decisions.

## You are given

- {SELLER}: what the user sells, to whom.
- {PROSPECT}: name, company, role, sector, how the meeting arose, first or repeat.
- {PROSPECT_BRIEF}: what they do, their sector's words, likely trigger.
- {STAGE}: pre-meeting or post-call.
- {SCORECARD}: Budget, Authority, Need, Timeline 0-25 each, signals tagged High, Medium, Low or Inferred with quoted sources, MEDDIC completeness, grade.
- {ACTION}: accept, discovery only, or short call.
- {GAPS}: elements lacking buyer evidence, most valuable first.
- {KEY_QUESTION}, {TRAP}.
- Post-call only: {ANSWERS}, {NEXT_STEP}, {BIGGEST_RISK}, {EXIT_POINTER} ("none" when qualified).

## Produce

Pre-meeting, three files.

`scorecard.md`: table, rows Budget, Authority, Need, Timeline: score, signals with tag and quoted source, gap. Under it: completeness, grade, {ACTION} verbatim.

`discovery-questions.md`: exactly ten questions, ordered 1-2 rapport and context, 3-4 current situation, 5-6 pain, 7-8 impact, 9-10 future state and decision process. Each: Question, Purpose, Expected Response, Follow-Up, Listen For. Spend 5-10 on {GAPS}. Repeat meeting: question 1 is "What has changed since we last talked?" Open 3, 5, 7 and 9 with "Can you walk me through" or "Can you help me understand".

Bank, reworded into the sector's language:
- Pain, impact: "How is [pain] showing up in the business today?", "Where is that now, and where do you want it?", "Have you done the maths on what ideal state is worth?", "What have you tried so far?"
- Urgency: "What happens if this is not addressed now?", "Fast forward 90 days: is it acceptable if this is still unsolved?" Never "What is your timeline?"
- Decision process: "What was the process last time you bought [similar service]?", "Who else has a strong view on this?"
- Economic buyer: "Who ultimately signs off, and what number are they responsible for?"
- Champion: "What objections will we face?", "What is missing from the business case for [decision maker]?"
- Paper process, repeat meetings only: "Who actually signs, is there more than one?", "Is there time left in the quarter?"
- Budget, after impact only: "What do you spend on this today?"

`cheat-sheet.md`, under 120 words: grade, top three gaps, {KEY_QUESTION}, {TRAP}, three Listen For signals.

Post-call. `qualification-note.md`, one page: scorecard rows re-scored from {ANSWERS} with the buyer's quoted words; completeness and grade; who else is involved and who signs, or "unknown"; {NEXT_STEP} and {BIGGEST_RISK} verbatim.

`graceful-exit.md`, only when {EXIT_POINTER} is not "none", under 120 words the user can send: thanks, plainly not the right fit now, {EXIT_POINTER} as something useful, a specific way back.

Current folder, plain Markdown, British English, no em dashes.

## Rules

- Only questions the salesperson genuinely wants answered; nothing leading, no pressure wording.
- Never "Tell me about your company", "How's business?", "What are your goals?", "What keeps you up at night?", "Are you the decision-maker?"
- No budget or authority question before question 7; no product or demo content.
- Expected Response and Listen For from {PROSPECT_BRIEF} only; "unknown" rather than invent.
- Every score, tag, source and decision exactly as given.

## Return

Print the stage, each file's full path, the question count, whether graceful-exit.md was written, then every file's contents.