Feature-to-Spike

┌─────────────────────┐
│ 1. User uses        │
│    launchpad-       │
│    toolkit skill    │ ◄────────── External founder or internal dog-food
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│ 2. Methodology      │
│    learning emerges │
│    (UX pattern,     │
│     flow improve-   │
│     ment, missing   │
│     feature, etc.)  │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│ 3. feature-to-spike │
│    transforms       │ ◄────────── THIS SKILL
│    learning into    │
│    SPIKE issue      │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│ 4. William evaluates│
│    SPIKE, decides   │
│    scope/priority   │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│ 5. DojoOS Launchpad │
│    feature shipped  │
│    (OR skill iterated│
│    if SPIKE rejected)│
└─────────────────────┘

Ver references/productization-workflow.md para detalle del workflow.

¿Qué es un good SPIKE para launchpad?

Un SPIKE útil para productizar en DojoOS Launchpad tiene 4 características:

1. Concrete pattern — no "William debería revisar UX" sino "founders en stage=Ideation con industry=fintech tienen 60% más engagement cuando el intake pregunta traction signals ANTES de competitive landscape"
2. Dogfood evidence — claim fundamentado en ≥1 sesión de dog-food con artefactos generados (startup-profile.md, etc.)
3. Productization hypothesis — una idea concreta de cómo productizarlo en DojoOS (UI flow, data model change, algorithm tweak...)
4. Acceptance criteria — cómo William puede validar si la hipótesis funciona post-implementation

SPIKES que NO cumplen estas 4 son mejores como notes o user feedback, no como SPIKE formal.

Flujo del skill

Paso 1 — Identificar el learning

FTS-1: "¿Qué aprendiste del dog-food que podría valer productizarse en DojoOS Launchpad? Describílo en 1-2 oraciones (el 'concrete pattern')."

Validar que el pattern es concreto y specific. Si vago → re-prompt: "¿Podés dar un ejemplo concreto del pattern en acción?"

Paso 2 — Evidence gathering

FTS-2: "¿Qué artefactos o sesiones de dog-food soportan este learning?

• Skill(s) usado(s): [startup-intake / cap-table-builder / etc.]
• Artifact(s) generado(s): [path a .md archivos]
• Observación específica: [qué viste]
• Startup(s) involucradas: [nombres o anonymizado si confidential]"

Paso 3 — Productization hypothesis

FTS-3: "¿Cómo te imaginás productizarlo en DojoOS Launchpad? Describí:

• Qué component(s) se tocarían: [UI / API / data model / matching algorithm]
• Cambio concreto: [ej: 'reordenar sections del intake form', 'agregar field X al Startup Profile schema', 'incluir Y en co-founder matching weight']
• Alternativas consideradas: [brief]"

Paso 4 — Acceptance criteria

FTS-4: "¿Cómo validaría William si funcionó post-implementation? Definí ≥2 criterios measurables:

• Metric: [ej: intake completion rate, profile quality score, matching accuracy]
• Threshold: [cuánto mejora vs baseline]
• Timeline: [cuándo medir]"

Paso 5 — Priority + scope

FTS-5: "En tu opinión:

• Priority (Urgent/High/Normal/Low): [default = Normal]
• Est. scope (S/M/L/XL): [S = 1 session, L = 3-5 sessions]
• Blocking dependencies: [ej: requires DojoOS API by @garbanzo, blocked on X]"

Paso 6 — Generate SPIKE draft

FTS-6: Generar ./launchpad/spikes/YYYY-MM-DD-{spike-slug}.md con el template de output.

FTS-7: "¿Filear el SPIKE en Linear ahora, o revisás el draft primero?

• Opción A: File ahora (requiere Linear MCP o gh CLI con Linear integration)
• Opción B: Solo genero el draft, vos lo fileás manualmente cuando estés listo"

Output template

Generar ./launchpad/spikes/YYYY-MM-DD-{spike-slug}.md:

# SPIKE: [Short title — what to investigate/productize]

**Target assignee**: William Ugalde (Launchpad pillar owner)
**Suggested labels**: spike, launchpad, methodology-prototype
**Priority**: [Urgent / High / Normal / Low]
**Est. scope**: [S / M / L / XL]

---

## Context

Methodology learning from dog-food of `launchpad-toolkit`. This SPIKE proposes productizing a pattern validated via prototype in the plugin — reducing productization risk for DojoOS Launchpad.

**Concrete pattern observed**:

[1-2 sentences, specific + measurable]

---

## Dogfood evidence

### Skill(s) used

- [Skill name(s) from launchpad-toolkit]

### Artifact(s) generated

- [Path to `.md` artifact, ideally linked or attached]

### Observation

[What was observed — quote artifacts or specific outputs if helpful]

### Startup(s) involved

[Name or anonymized "Founder A, stage=MVP, industry=fintech"]

---

## Productization hypothesis

### Component(s) to change in DojoOS Launchpad

- [UI / API / data model / matching algorithm / etc.]

### Concrete change proposed

[ej: "Reorder intake form sections: Traction before Competitive to match observed dogfood pattern"]

### Alternatives considered

[Brief rejection rationale for other options]

---

## Acceptance criteria

- [ ] **Metric**: [ex: intake completion rate]
- [ ] **Threshold**: [ex: +15% vs baseline measured in 2-week cohort]
- [ ] **Timeline**: [ex: measure 30 days post-feature-ship]

- [ ] **Metric**: [second criterion]
- [ ] **Threshold**: [threshold]
- [ ] **Timeline**: [timeline]

---

## Dependencies / prereqs

- [List blocking dependencies — ex: "requires DojoOS API by @garbanzo" or "none"]

---

## Links

- **launchpad-toolkit SKILL**: [link to SKILL.md that generated the learning]
- **Artifact(s)**: [link to `.md` output(s)]
- **Related DOJ issues**: [parent SPIKE DOJ-3189, sibling issues if any]

---

Created by Claude Code via `launchpad-toolkit:feature-to-spike`, on behalf of @lapc506.

Integración con Linear

Vía Linear MCP (recomendado)

Si Linear MCP está configurado (mcp__linear-server__save_issue), el skill puede filear el SPIKE directamente:

mcp__linear-server__save_issue(
  team="DojoOS",
  project="Launchpad",
  parentId="DOJ-3189",           # Parent SPIKE for launchpad-toolkit
  assignee="[email protected]",
  title="SPIKE: [title]",
  description="[generated body]",
  labels=["Spike", "Explore", "M"],
  priority=3,
  state="Triage"                  # Let William move to In Progress when ready
)

Vía gh + Linear sync

Si gh CLI está disponible pero Linear MCP no, generar issue body + prompt user para filear manualmente con el slug andres/doj-XXXX-{slug}.

Vía copy-paste

Si ningún tooling disponible, output puro markdown para que usuario copie-pega en Linear web UI.

Principios clave

• NO crear issues sin confirmación explícita — default es generar draft, filear solo si user dice "sí"
• William como default assignee (ID Linear: 8f14370d-3602-49e3-81f2-eeb05b965687; email: [email protected])
• Parent issue = DOJ-3189 por default — mantiene trazabilidad con el tracking SPIKE del plugin
• Labels consistentes: spike, launchpad, methodology-prototype — William filtra por estos
• No escalate prematurely: si el learning es débil, sugerir "keep dogfooding más antes de SPIKE"

Anti-patterns

• SPIKEs vagos tipo "mejorar UX" — rechazar, pedir specific pattern
• SPIKEs sin acceptance criteria — no hay forma de validar post-implementation
• Crear issues bypaseando William → viola ownership del pillar (ver feedback memory)
• Usar SPIKE format para feature requests — si es clear feature (no investigation), debería ser feature issue, no SPIKE

Integración con otras skills

• Input usuario de cualquier skill de launchpad-toolkit → si durante uso detectan un productization candidate, invocar feature-to-spike
• startup-intake: learnings sobre preguntas que funcionan mejor
• cap-table-builder (v0.2): learnings sobre edge cases del vesting calculator
• cofounder-matching (v0.2): learnings sobre weighting del algoritmo

Recursos

• DojoOS Launchpad pillar ownership (Slack 2026-04-10 #C0AKTN24C91)
• Parent tracking SPIKE: DOJ-3189
• Productization workflow doc: references/productization-workflow.md