What it does
This skill takes whatever evidence a deal has produced, call transcripts, email threads, CRM notes, the previous review, and scores it against a qualification framework, producing a MEDDICC scorecard by default or a BANT, SPICED, or custom version if you tell it. The job it replaces is the ten minutes before pipeline review where a rep fills in seven fields from memory and the manager nods, because neither of them has the evidence in front of them. The skill reads the evidence first and scores second, and it will not give a field more than a 1 unless the buyer said it or did it.
A bad scorecard is a set of green boxes. Every field reads 'yes', the total looks healthy, and the deal slips a quarter because nobody checked whether the economic buyer had ever been on a call. A good scorecard has a scale with teeth (0 no evidence, 1 rep asserts, 2 buyer confirmed, 3 buyer confirmed and acted), one line of quoted evidence per field with its source and date, and a gaps section ordered by how much each gap threatens the close date. The difference is the evidence column. Once every score has to point at a quote, optimism has nowhere to hide.
It beats the manual version because it is consistent and because it remembers. Consistency: two reps' deals get scored on the same scale by the same reader, so a manager can compare a 15/21 to a 15/21 and trust it, and a forecast built on those totals starts to mean something. Memory: hand it the prior scorecard and it shows movement per field and flags fields that have not moved across two or more reviews, which is the single most reliable stall signal in a pipeline and the one humans miss because each review feels like a fresh conversation. It also forces the one habit reps skip, naming a single next action with an owner and a date, instead of 'keep pushing'.
Where it does not help: it cannot score what nobody wrote down. If the evidence is a rep's memory of a call that was not recorded, everything caps at 1, correctly, and the fix is to record the next call rather than to argue with the scale. It does not replace the conversation with the buyer; it tells you which conversation to have next and what you need to hear on it. It does not know your deal history, so a first scorecard has no movement section, and it does not know your stage definitions unless you paste them. Treat it as the referee in deal review, not the coach, and not the forecast.
The MEDDICC scorecard template
This is the blank the skill fills. Copy it into your deal review doc or paste it as the output format when you run the skill. The scoring scale is the part that matters: 0 means no evidence, 1 means the rep says so, 2 means the buyer said so, 3 means the buyer said so and did something about it. Hold that line and the total starts predicting close dates.
DEAL SCORECARD, {{DEAL}}
Stage: {{stage}} Amount: {{amount}} Close: {{date}} Reviewed: {{date}}
Scale: 0 no evidence / 1 rep asserts / 2 buyer confirmed / 3 confirmed and documented or acted on
METRICS [{{0-3}}] Evidence: {{quote or action}} ({{source, date}})
ECONOMIC BUYER [{{0-3}}] Evidence: {{...}}
DECISION CRITERIA [{{0-3}}] Evidence: {{...}}
DECISION PROCESS [{{0-3}}] Evidence: {{...}}
IDENTIFY PAIN [{{0-3}}] Evidence: {{...}}
CHAMPION [{{0-3}}] Evidence: {{...}}
COMPETITION [{{0-3}}] Evidence: {{...}}
TOTAL: {{n}}/21 Prior review: {{n}}/21
GAPS (largest threat to the close date first)
1. {{field}}: {{what is missing}}
2. {{field}}: {{what is missing}}
STUCK (unchanged 2+ reviews): {{fields}}
NEXT ACTION: {{one action}} Owner: {{name}} By: {{date}}
VERDICT: {{Commit / Best Case / Pipeline / At Risk}}, driven by {{field}}Swap the seven rows for BANT or SPICED fields if that is what you run; the scale and the gaps section stay the same. A scorecard where every field is a 3 and the deal still slipped means the scale was too generous, not that the framework failed.
Inputs & outputs
Inputs
- Deal name, stage, amount, and target close date (required)
- Framework: MEDDICC (default), BANT, SPICED, or a custom list of fields
- Evidence: call notes or transcripts, email threads, CRM notes, a forwarded one-pager, anything the buyer said or did
- Optional: the prior scorecard, so the skill can show movement and flag stuck fields
- Optional: what you sell and to whom, so Metrics and Pain are judged against a real value story
Outputs
- A one-page scorecard with a 0-3 score per framework field and one line of quoted evidence with source and date for each
- A total against the maximum, with the prior total beside it
- Gaps ranked by threat to the close date, and a Stuck list of fields unchanged across two or more reviews
- One next action with an owner and a date, not a list
- A verdict (Commit, Best Case, Pipeline, At Risk) and the single field that drives it
How to set it up
Decide where the skill lives
Two setups, both work. (1) In claude.ai, create a Project called "Deal Review", open Project settings, and paste the SKILL.md body below into custom instructions; add one line naming your framework and what you sell. Every chat in that Project scores deals. (2) In Claude Code or Claude Desktop skills, save the content as skills/deal-qualification-scorer/SKILL.md and invoke it by name. The Project route suits reps; the folder route suits a RevOps owner who wants to run every open deal in a loop.
Feed it evidence, not opinions, for the MEDDICC scorecard
Paste the actual material: the transcript from your call recorder, the email where the champion listed their criteria, the CRM notes. Do not summarize it first; the skill is looking for buyer words and buyer actions, and a rep summary launders assertions into facts. If you have the prior scorecard, paste that too so the movement section works. If the framework is not MEDDICC, say so up front: "Score this on SPICED" or paste your custom fields.
Read the gaps before the total
The total is for the manager. The rep should read the gaps section first, because it is already ordered by threat to the close date and the top item is the call to book this week. A Decision Process at 0 with six weeks to close is a bigger problem than a Competition at 1, and the skill has already done that ranking. Argue with a score if you have evidence it missed, then paste the evidence and re-run; do not just bump the number.
Re-score at every review and watch the Stuck list
Run it on the same deal before each pipeline review with the last scorecard attached. The movement section shows what actually progressed, and the Stuck list catches the field that has read 1 for three weeks while the deal sat in Commit. Pair it with the discovery-call-prep skill: the framework tags on the prep questions are the same fields this skill scores, so a well-run discovery call fills the scorecard directly.
The SKILL.md
Save this as SKILL.md in a folder named deal-qualification-scorer, or paste the body into a Claude Project. Then send it a deal and its evidence.
---
name: deal-qualification-scorer
description: Use when a rep or manager needs to score a live deal against a qualification framework (MEDDICC default; BANT, SPICED, or custom). Takes call notes, transcripts, emails, and CRM fields and produces a scorecard with a 0-3 score per field, the evidence behind each, the gaps, and the next action. Never scores on a rep's assertion alone.
---
# Deal Qualification Scorer
You score one deal against one qualification framework and tell the rep what is missing. Output a scorecard a manager can read in 90 seconds. Be strict: the scorecard exists to stop optimism reaching the forecast.
## Inputs
- Deal name, stage, amount, and close date (required).
- Framework: MEDDICC (default), BANT, SPICED, or a custom list of fields.
- Evidence: call notes or transcripts, emails, CRM notes, the prior scorecard if any.
- Optional: what the user sells and to whom, so Pain and Metrics are judged against a real value story.
## Method
1. Read all evidence before scoring. Note the date and source of each item.
2. For each field, find the strongest evidence. Quote or paraphrase it and name the source (call date, email from the champion, CRM note).
3. Score each field 0-3: 0 = no evidence; 1 = the rep asserts it but the buyer has not confirmed it; 2 = the buyer confirmed it in their own words; 3 = confirmed AND documented or acted on (a named number, a signed MAP step, an intro made).
4. "They love it" or "budget is fine" from the rep is a 1 at most. Only buyer words or actions reach 2 or 3.
5. Total the scores. Flag every 0 or 1 as a gap, ordered by threat to the close date.
6. If a prior scorecard is given, show movement and flag fields unchanged across 2+ reviews.
7. Name the single next action that fills the largest gap, with owner and date.
8. Verdict: Commit, Best Case, Pipeline, or At Risk, and the field that drives it.
## Hard rules
- Evidence or zero. No 2 or 3 without a buyer quote, buyer action, or a document.
- Never invent evidence, names, numbers, or dates. Unjudgeable field = "no evidence", score 0.
- Do not soften. A scorecard that reads well and closes late has failed.
- One page. One line of evidence per field.
## Output template
**Deal Scorecard: <Deal>, <today's date>, <framework>**
Stage | Amount | Close | Sources: <list>
| Field | Score | Evidence (source) |
|---|---|---|
| <Field> | <0-3> | <quote or paraphrase> (<source>) |
**Total: <n>/<max>** (prior: <n> or "first score")
**Gaps, largest threat first**
1. <Field>: <what is missing, why it threatens the close date>
2. <Field>: <...>
**Movement since last review**
- <Field> <old> -> <new>: <why>
- Stuck: <fields unchanged across 2+ reviews>
**Next action**
<One action, owner, date.>
**Verdict:** <Commit | Best Case | Pipeline | At Risk>, driven by <field>.
Run it from the terminal
Example run
An AE scores a proposal-stage logistics deal for Monday's pipeline review, pasting two call transcripts and a champion email.
Prompt: deal-qualification-scorer. Deal: Northwind Freight, Proposal, $84k ARR, close 2026-10-15. Framework: MEDDICC. Evidence: transcript 2026-08-27, notes 2026-09-02, champion email 2026-09-03, CRM notes. We sell dispatch automation to mid-size freight brokers.
Output:
Deal Scorecard: Northwind Freight, 2026-09-04, MEDDICC
Stage: Proposal | Amount: $84k ARR | Close: 2026-10-15 | Sources: call 2026-08-27 (transcript), call 2026-09-02 (notes), champion email 2026-09-03, CRM notes
| Field | Score | Evidence (source) |
|---|---|---|
| Metrics | 2 | "We lose about 6 hours a week per dispatcher to manual rate lookups" (call 2026-08-27) |
| Economic Buyer | 1 | Rep note says "VP Ops owns budget"; VP Ops has not been on a call (CRM) |
| Decision Criteria | 2 | Champion listed TMS integration, SSO, and onboarding under 30 days (email 2026-09-03) |
| Decision Process | 0 | No evidence of approval steps, security review, or legal timeline |
| Identify Pain | 3 | Dispatcher turnover tied to lookup workload; champion cited two resignations (call 2026-08-27) |
| Champion | 2 | Ops manager set up both calls and asked for a one-pager to forward (email 2026-09-03) |
| Competition | 1 | Rep believes "no one else is in it"; buyer has not said so (CRM) |
Total: 11/21 (first score)
Gaps, largest threat first
1. Decision Process: 0. Six weeks to close with no known approval path is the biggest risk to 2026-10-15.
2. Economic Buyer: 1. The VP Ops has never heard the pitch; the champion is carrying it alone.
3. Competition: 1. "No one else" is an assumption; ask directly.
Movement since last review
- First score, no comparison.
Next action
Rep asks the champion on Monday 2026-09-08 for a 20-minute call with the VP Ops, and for the steps between "yes" and a signed order form, with names.
Verdict: Pipeline, driven by Decision Process.Related plays
- Turn Discovery Into a Business Case and Mutual Action Plan Template →What to build once the scorecard says the deal is real: the business case and the dated plan the champion sells internally.
- Discovery Call Questions, Generated for Every Account →The prep skill tags every question to the same framework fields, so a good discovery call fills this scorecard.
- Pipeline Digest: A Weekly Claude Routine That Reads Every Deal Before You Do →Run the scorer across every open deal on a schedule and roll the verdicts into one digest.
- Claude + Salesforce integration guide →Pulling the evidence (notes, emails, stage history) from the CRM over MCP instead of pasting it.