STATUS — Question Set Builder (free skill)
Current: v1.1.0, 2026-08-11. Owner: Cole.
What this is
The free skill in front of the $97 Blueprint. Reads one prospect call and returns the owner's own question set, marked up — the questions they ask every time, the ones that got a dead answer with replacements, every place they answered for the prospect, and the talk ratio. Two input paths: a pasted transcript, or the questions typed from memory.
Scoring engine is Kathryn's 4-Move Question Framework™ (M1–M4).
Version history
| Version | Date | What |
|---|---|---|
| 1.0.0 | 2026-08-10 | First build (Cole). Superseded — do not restore. |
| — | 2026-08-11 | Kathryn's own v1.0.0, produced independently. Not landed as a file; its rules are merged into 1.1.0. |
| 1.1.0 | 2026-08-11 | Current. Merge — Cole's framework spine, Kathryn's integrity rules. |
What 1.1.0 changed, and why
Merged in from Kathryn's version: the counts/doesn't-count list with brevity is not death; the no-diagnosis passage; never characterize the prospect; the speaker-label boundary rule; honest-measurement language; the 800-word body cap with a stated cut order.
Kept from 1.0.0: M1–M4 scoring, the standing/one-off split with the total count, greeting exclusion reported, both input paths first class, never skip the number, The One Change.
New — the Framework Boundary section. Resolves the collision between the two: naming the move a question broke is a claim about the question and stays; explaining why the listener answered as they did is a claim about them and goes.
Defects fixed in 1.0.0:
- An invented 30/70 benchmark printed with a "Gap: X points over" verdict. Removed from the spec, the output template, and the Notion page. It told Kathryn she was failing a standard that existed nowhere, on a paid session she was hired to talk in.
- Diagnosis lines. Every "Why it died" row is now "Move broken" and takes the question as its subject.
- A word-count proxy for dead answers ("under about ten words") that caught complete short answers.
- Section numbering against the output template; metadata
updated→date.
Defect found and fixed during the retest: the merged recall path computed a share against a denominator that included "the modeled time the user's own questions consume in the asking" — a constant defined nowhere in the file. Uncomputable, and at 5/8/12 seconds per question the same set returned 85%, 77%, 70%. The recall path now reports projected minutes of their voice, no percentage. Movement is preserved: swapping 3 closed questions for open on an 8-question set goes 3.7 → 6.2 minutes.
Verified 2026-08-11
| Check | Result |
|---|---|
| Retest against the 7/27 call | Dead answers 9 → 5 → 4 counted + 1 borderline. Three of four fall out mechanically on the prospect's own words. |
| Talk ratio, third independent pass | 44/56 first fifteen, 58/42 whole call. Matches both prior passes. |
| Diagnosis lines | Zero. The two worst from 1.0.0 are quoted in the retest so they can't return. |
| Body word cap | Full audit rendered at 740 words against the 800 cap. Appendix (~486 words) sits outside it. |
| Recall path math | Run. Denominator removed — see above. |
| Hosted file | Uploaded, invalidate: true. Live URL serves 26,069 bytes, v1.1.0, byte-identical to local (diff clean). The old versioned URL also resolves to 1.1.0. |
| US spelling | Swept. |
Files
| File | What |
|---|---|
01-question-set-builder-skill.md | The skill. v1.1.0. |
01-question-set-builder-quick-start.md | Buyer-facing quick start. |
01-question-set-builder-delivery-page-cyp.html | Convertri-route delivery page, CYP brand. |
test-output-question-set-builder-2026-08-10.md | 1.0.0 test run. Historical. |
test-output-question-set-builder-v1.1.0-2026-08-11.md | The retest. Internal — contains client-identifiable content, never ships. |
Delivery
- Hosted:
https://res.cloudinary.com/dhldvrfdh/raw/upload/question-set-builder-skill.md— unversioned, always current. - Notion delivery page: https://app.notion.com/p/3b9a879ece9d81989c31ec5b6cf0e79c — NOT PUBLISHED. Kathryn publishes: Share → Publish to web, plus Allow duplicating as template.
Open
- The name. Built as Question Set Builder. Question Set Builder and The Dead Question Finder are still live options. A rename touches six places, ~10 minutes. Blocks Sloane's copy, since the email carries
[SKILL NAME]. - Send copy — Sloane's. Kathryn's two-email sequence is the copy; needs her voice pass and
copy-gate. - Incognito duplicate test — can't run until published.