Skill-to-Marketplace Proof Pack

Complete demonstration

This is the same-case worked artifact shipped with the Skill. It preserves the evidence boundaries and does not claim publication or commercial results.

# Demonstration Output — Marketplace Proof Pack

Fictional data, not supplied by the user. No upload, pricing save, submission,
review, publication, install, or purchase occurred.

## Artifact identity

- Runtime: fictional `meeting-decision-extractor`
- Buyer job: turn supplied meeting notes into a decision register.
- Evidence available: structural parse and one fictional cold-run output.
- Evidence unavailable: independent runtime, cross-model, marketplace, and buyer
  evidence.
- Price position: $5 low-ticket hypothesis, not an optimal-price finding.

## Listing draft

### Title

Meeting Decision Extractor — Owners, Due Dates & Open Questions

### Short description

Turn messy meeting notes into a clear decision register with named owners, due
dates, unresolved questions, and an explicit missing-evidence section.

### Buyer problem

The meeting ends, but the notes still mix discussion, decisions, requests, and
ideas. Owners disappear into paragraphs and unresolved questions look complete.

### Concrete outcome

Paste supplied notes and receive one structured register separating:

- confirmed decisions;
- actions and named owners;
- due dates explicitly present in the notes;
- open questions;
- missing owner or date fields.

### How it works

1. Separates statements by decision state.
2. Preserves exact owner and date evidence.
3. Marks missing fields instead of guessing.
4. Returns a compact register ready for human review.

### Honest boundary

It does not record meetings, identify speakers not named in the notes, contact
participants, update calendars, or prove that an action was completed.

### CTA

Install once and turn your next note dump into a reviewable decision register.

## Worked buyer request

```text

**Use these notes**
- We agreed to ship the onboarding copy after legal review.
- Nina owns the revised draft for Friday.
- The pricing page question is still open.
- Someone should confirm the analytics event names.

Worked result

MEETING DECISION REGISTER — Demonstration Output

DECISION D01
**State:** CONDITIONAL
**Decision:** Ship the onboarding copy after legal review.
**Owner:** UNKNOWN
**Due date:** UNKNOWN
**Dependency:** Legal review

ACTION A01
**Action:** Prepare the revised onboarding copy.
**Owner:** Nina
**Due date:** Friday — exact calendar date not supplied
**Dependency:** D01

OPEN QUESTION Q01
**Question:** What decision is required for the pricing page?
**Owner:** UNKNOWN

OPEN QUESTION Q02
**Question:** Who will confirm the analytics event names?
**Owner:** UNKNOWN

**VERDICT:** READY FOR HUMAN REVIEW — no calendar or message action occurred

Three-shot proof sequence

Screenshot 1 — Before / Input

• Headline: Your meeting notes are not a decision system.
• Visual: the exact four-line buyer request with discussion and ownership mixed.
• Evidence badge: FICTIONAL DEMONSTRATION INPUT.
• Proves: recognizable problem and low-friction input.
• Alt text: Four fictional meeting-note bullets containing a conditional
  decision, one named owner, and two unresolved questions.

Screenshot 2 — After / Output

• Headline: Decisions, owners, dates, and unknowns—separated.
• Visual: D01, A01, Q01, and Q02 from the same result.
• Evidence badge: SAME-CASE DEMONSTRATION OUTPUT.
• Proves: the promised structure and visible missing-data behavior.
• Alt text: Structured fictional decision register with one conditional
  decision, one action assigned to Nina, and two open questions.

Screenshot 3 — Mechanism / Boundary

• Headline: It marks what the notes do not prove.
• Visual: dependency line, unknown owner/date fields, and external-action boundary.
• Evidence badge: STRUCTURAL + FICTIONAL COLD RUN ONLY.
• Proves: evidence discipline and scope.
• Alt text: Close view of missing fields and the statement that no calendar or
  participant contact action occurred.

Buyer-guide outline

1. What problem the Skill owns.
2. Quick start with one note block.
3. Supported inputs and maximum input count.
4. Decision, action, and question states.
5. How unknown owners and dates are handled.
6. Complete worked example.
7. What the Skill does not do.
8. Support path and safe reporting information.

Claim cards

┌─ CLAIM CARD C01 ────────────────────────────────────────┐
│ Claim: turns supplied notes into a decision register    │
│ Support: runtime instruction + fictional cold run       │
│ Status: SUPPORTED AS DEMONSTRATION                      │

│ Dependency: Worked result -> Screenshot 2 │
└──────────────────────────────────────────────────────────┘

┌─ CLAIM CARD C02 ────────────────────────────────────────┐
│ Claim: works across AI agents                            │
│ Support: no cross-model evidence                         │
│ Status: REMOVE                                           │

│ Dependency: compatibility field -> FAQ │
└──────────────────────────────────────────────────────────┘

┌─ CLAIM CARD C03 ────────────────────────────────────────┐
│ Claim: saves hours                                       │
│ Support: no time study                                   │
│ Status: REMOVE                                           │
│ Dependency: title and short description                 │
└──────────────────────────────────────────────────────────┘

Release blockers

• Platform field limits and current policy: UNVERIFIED.
• Independent runtime execution: NOT RUN.
• Cross-model behavior: NOT RUN.
• Final package security preflight: NOT RUN in this fictional demonstration.
• Visual assets: specified, not rendered.

Signal screen

✓ Runtime owner: fictional completed Skill
✓ Listing/demo alignment: same buyer job and same case
✓ Proof state: structural parse + fictional cold run only
✓ External state: no upload or publication claim
✗ Remaining gate: current marketplace constraints and independent runtime

**VERDICT:** DEMONSTRATION — RUNTIME EVIDENCE NEEDED

Release dependency gate

RELEASE DEPENDENCY GATE v1.1
**Runtime owner:** PRESENT — fictional Meeting Decision Extractor
**Worked request/result pair:** PRESENT — demonstration only
**Proof shots:** 3/3 SPECIFIED — not rendered screenshots
**Validation evidence:** STRUCTURAL FIXTURE CHECK ONLY
**External state:** LOCAL DRAFT — no upload, submission, approval, sale, or purchase
**Release verdict:** HOLD FOR CREATOR REVIEW

Next action: run the completed runtime independently on the worked request before
using any runtime-tested claim.

**Next step:** review the evidence ledger and release blockers before using the asset externally.