Mpp Evaluator

MPP Evaluator

What Is MPP?

Minimum Proud Product — the minimum version you'd be genuinely proud to show to someone you deeply respect. Not just functional. Not just shipped. Something where a user feels something.

You've hit MPP when:

• You'd be embarrassed not to show it to someone you admire
• At least one user has described it to someone else without being asked
• The product creates a feeling, not just a function
• There's evidence in the details that someone cared deeply

Quick Start

Human-First Response

Default: answer in chat first with Verdict, Do this now, and Details saved. Keep chat under 150 words unless asked; do not paste the full artifact. Write/update the requested .md file as the durable record for agents.

Say: "Evaluate my MPP" or "Run the MPP scorecard" or "Am I proud of this?"

Output: mpp-scorecard.md — five criteria scored by independent agents, composite score, one gap named.

---

Parallel Execution

Five independent criteria. Five parallel agents. All five can be scored simultaneously — there's no dependency between them.

Before spawning agents, read:

• founder-context.md — current stage, north star, last MPP score
• cycles/ — the last 3 cycle records for behavioral evidence

Spawn these 5 agents simultaneously:

Agent 1 — Pride Scorer Reads: founder-context.md + last cycle + any demo/share mentions in cycles Task: Score the Pride Test (1–10): "Would the founder show this without apologizing?" Returns: Score + 2-sentence evidence statement + the specific thing blocking a higher score

Agent 2 — Recommendation Scorer Reads: founder-context.md + all cycles for mentions of sharing/referrals Task: Score the Recommendation Test (1–10): "Has any user described it to someone else unprompted?" Returns: Score + specific evidence (quotes from cycles if any, absence if none) + what would trigger sharing

Agent 3 — Emotion Scorer Reads: founder-context.md + cycle records for user behavior descriptions Task: Score the Emotion Test (1–10): "What do users feel after the core task?" Returns: Score + evidence from user feedback in cycles + what the product currently makes users feel vs. what it should

Agent 4 — Craft Scorer Reads: founder-context.md + any UX or design notes in cycles Task: Score the Craft Test (1–10): "Is there evidence someone cared about the details?" Returns: Score + specific gaps identified (error messages, empty states, copy quality) + 1 highest-leverage craft fix

Agent 5 — Grief Scorer Reads: founder-context.md + cycle records for retention and engagement signals Task: Score the Grief Test (1–10): "How would the best user react if this disappeared tomorrow?" Returns: Score + evidence (retention data, user quotes) + what would need to be true for a higher grief score

Wait for all 5 agents to return. The orchestrator (main agent) synthesizes.

---

Synthesis (Orchestrator Only)

After all 5 scores return:

1. Calculate composite: Average of 5 criterion scores.

2. Find the gap: Which single criterion, if improved by 2+ points, would move the composite most? That's the gap to name.

3. Write mpp-scorecard.md (orchestrator is the only writer):

# MPP Scorecard — [YYYY-MM-DD]

## Scores
| Criterion | Score | Key Evidence |
|-----------|-------|-------------|
| Pride Test | [X]/10 | [Agent 1's evidence statement] |
| Recommendation Test | [X]/10 | [Agent 2's evidence statement] |
| Emotion Test | [X]/10 | [Agent 3's evidence statement] |
| Craft Test | [X]/10 | [Agent 4's evidence statement] |
| Grief Test | [X]/10 | [Agent 5's evidence statement] |

**Composite: [X.X]/10**

## Interpretation
[Score range interpretation — see criteria reference]

## What MPP Looks Like for [Your Product]
[Specific description of what THIS product needs to be to earn MPP — not generic]

## The One Gap
[The criterion holding back the score most + the specific fix]

## What Changes When You Close This Gap
[One sentence — what happens to the product and users if this gap closes]

4. Update founder-context.md: Set mpp-score to the new composite.

---

The Five Criteria (Quick Reference)

For the deep criteria guide, see [mpp-criteria.md](mpp-criteria.md).

| Criterion | The Question | 1–3 | 4–6 | 7–9 | 10 | |-----------|-------------|-----|-----|-----|-----| | Pride | Show without apologizing? | Avoid showing | Show with caveats | Show without apology | Excited to share | | Recommendation | Unprompted sharing? | None | 1–2 instances | Regular pattern | Measurable word-of-mouth | | Emotion | What users feel after core task? | Nothing | Mild satisfaction | Clear positive feeling | Delight | | Craft | Evidence someone cared? | Rough, placeholder | Mostly functional | Details considered | Every touchpoint intentional | | Grief | React if it disappeared? | Would shrug | Mild inconvenience | Genuinely frustrated | Devastated |

Composite ≥ 7.0 = MPP achieved. Below 7.0: fix the lowest criterion first.

---

Sequential Fallback (Codex / OpenCode)

If your agent doesn't support parallel subagents, run each criterion assessment in sequence:

1. Pride Test → score + evidence
2. Recommendation Test → score + evidence
3. Emotion Test → score + evidence
4. Craft Test → score + evidence
5. Grief Test → score + evidence
6. Synthesize → write mpp-scorecard.md

Same output. ~4× longer.

---

Related Skills

• build-cycle uses this skill's criteria in Phase 3 (abbreviated in-cycle version)
• failure-navigator — if MPP score is flat across 3+ cycles, run this next
• co-founder (Partner phase) — uses MPP trajectory as a key health signal

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: gvkhosla
• Source: gvkhosla/founder-skills
• License: MIT
• Homepage: https://fskills.xyz

Install and usage instructions live in the source repository linked above.