Ship 4 new skills + lead with Uri Levine

NEW SKILLS:
- /pivot — Eric Ries's 10 pivot types, evidence-driven; forces explicit
  kill criteria for the previous hypothesis; logs to brain/pivots.md
- /story — Uri Levine's narrative arc constructor (origin → why-you →
  journey → vision → ask); every claim cites a brain entry
- /pitch — renders /story into a 10-slide deck; refuses fabricated TAM,
  vanity metrics, unsupported quotes; flags weak slides honestly
- /pmf-check — triangulates Sean Ellis test + retention cohort + believer
  density; returns Strong PMF / PMF / Pre-PMF / PMF Mirage

URI LEVINE PROMINENCE:
- New hero callout: "An executable version of Uri Levine's Fall in Love
  with the Problem"
- Methodology section now leads with Uri (more space than others)
- Closing "With deep respect to Uri Levine" section before License

OTHER:
- Brain templates: pivots.md, story.md, metrics/, pitch/
- Updated CLAUDE.md to reflect 10 skills, grouped by purpose
- Tests now cover all 10 skills (134/134 passing)

Author: nocoder26

.claude/skills/pitch/SKILL.md

@@ -0,0 +1,126 @@
+---
+name: pitch
+description: Render brain/story.md into a 10-slide pitch deck where every claim cites a brain entry. Refuses to fabricate TAM, vanity metrics, or unsupported quotes. Flags weak slides honestly so the founder fixes them before any VC sees the deck.
+---
+
+# /pitch
+
+Read `brain/story.md` (must exist — run `/story` first if not). Read `brain/problem.md`, `brain/believers/index.md`, `brain/interviews/`, `brain/decisions.md`, `brain/runway.md`, `brain/metrics/` (if exists).
+
+If `brain/story.md` is empty: refuse. *"No story captured. Run /story first — the deck is downstream."*
+
+## The deck (10 slides, YC-style)
+
+Every slide cites specific brain entries. **Every. Single. Slide.**
+
+### Slide 1 — Title
+- Project name
+- One-line tagline (from `brain/story.md` "vision" condensed)
+- Founder name + handle
+
+### Slide 2 — Problem
+- 1-3 sentences from `brain/problem.md`
+- ONE verbatim quote from a believer (cite interview file)
+- **DO NOT** include "X% of Y have this problem" stats unless `brain/evidence/` supports it
+
+### Slide 3 — Customer
+- Segment from `brain/believers/index.md` (top concentration)
+- Believer profile: pain, behavior, willingness to pay
+- Cite N believers identified
+
+### Slide 4 — Solution
+- **Only** what's been validated
+- For each feature: which believer asked for it (cite interview file)
+- Unbuilt features marked "v1.5 (planned)" — don't pretend they exist
+
+### Slide 5 — Traction
+- Real numbers from `brain/metrics/` and `brain/believers/`
+- Sean Ellis % if measured (from `/pmf-check`)
+- Retention curve if measured
+- **DO NOT** include vanity metrics (downloads, signups, MAU without retention)
+- If traction is thin: write *"Pre-traction. N believers identified, MVP launching wk Y."* — admit it openly
+
+### Slide 6 — Why now
+- Founder fills. Claude prompt: *"What changed in the world that makes this possible/needed now? Be specific. Don't say 'AI is hot.'"*
+- If founder can't answer, mark slide WEAK in the assessment
+
+### Slide 7 — Team
+- Founder fills. Highlight founder-market fit from `brain/why.md`
+- Co-founders: name, role, why-them
+
+### Slide 8 — Business model
+- Pricing from `brain/decisions.md` if logged
+- Unit economics if `/default-alive` has been run (cite `brain/runway.md`)
+- If untested: mark *"Hypothesis — not yet validated."*
+
+### Slide 9 — Vision
+- 1-3 sentences from `brain/story.md` "vision"
+- The changed world if this works at scale
+
+### Slide 10 — Ask
+- From `brain/story.md` "ask"
+- Calibrated to `brain/runway.md`
+- Specific number, specific milestone
+
+## Output: brain/pitch/deck.md
+
+Markdown format renderable into Keynote / Pitch.com / Slidev / Marp.
+
+For each slide, include a `## Citations` block listing the brain entries used.
+
+```markdown
+# <Project name>
+
+## Slide 1: Title
+<...>
+
+### Citations
+- brain/story.md "vision"
+
+---
+
+## Slide 2: Problem
+<...>
+
+> "<verbatim quote from believer>"
+> — <Name>, interview #<N>
+
+### Citations
+- brain/problem.md (current hypothesis)
+- brain/interviews/00X-<name>.md
+
+---
+[...]
+```
+
+## Honest deck assessment (mandatory after rendering)
+
+Surface weak slides explicitly. The founder needs to see the gaps before a VC does:
+
+```
+DECK ASSESSMENT:
+✅ Slide 2 (Problem):    strong — 3 verbatim quotes, specific segment
+✅ Slide 4 (Solution):   strong — every feature cites a believer
+⚠️  Slide 5 (Traction):   WEAK — only 12 believers, no retention data
+⚠️  Slide 6 (Why now):    WEAK — founder didn't fill in, currently generic
+✅ Slide 8 (Business):   adequate
+❌ Slide 9 (Vision):     generic — needs work before sending
+
+VCs will probe slides 5, 6, 9. Address these before any pitch meeting.
+```
+
+## Hard refusals
+
+- **Refuses** to write TAM if `brain/evidence/` doesn't support it. Will write *"TAM unknown — to be sized"* instead.
+- **Refuses** to write *"we are growing N% MoM"* without retention data in `brain/metrics/`.
+- **Refuses** to fabricate quotes. Period.
+- **Refuses** to write *"the market is $X billion"* without a cited source.
+- **Refuses** to call something a "platform" if `brain/decisions.md` doesn't show platform decisions.
+
+## Why this exists
+
+Most pitch decks are full of made-up numbers and unsupported claims. VCs detect this in 30 seconds. **The deck where every claim has a citation is rare and credible.**
+
+Uri Levine: *"You're not pitching the deck. You're pitching the story behind the deck."*
+
+This skill makes the deck a faithful render of the story — your story, your evidence, your receipts. When a VC asks *"how did you arrive at this number?"* you have the citation ready.

.claude/skills/pivot/SKILL.md

@@ -0,0 +1,111 @@
+---
+name: pivot
+description: Walk through Eric Ries's 10 pivot types using brain evidence to inform direction. Forces explicit kill criteria for the previous hypothesis before allowing a pivot. Logs to brain/pivots.md with full reasoning.
+---
+
+# /pivot
+
+Read `brain/problem.md`, `brain/believers/index.md`, `brain/interviews/`, `brain/decisions.md`, `brain/pivots.md` (if exists). Run `query_evidence()` for relevant signals.
+
+## When to invoke
+
+Founder is considering changing direction. Triggers:
+- Believer density <30% after 5+ interviews
+- Retention curve approaching zero
+- Multiple "this isn't quite my problem" signals
+- Founder explicitly says *"I'm thinking about pivoting"*
+
+## Step 1 — Surface what's actually wrong
+
+Before naming a pivot type, name the failure. Read brain/ and ask:
+- What did the original hypothesis predict?
+- What did the evidence show?
+- Where exactly did they diverge?
+
+If the founder hasn't done a `/problem` refresh recently, suggest it before pivoting. Often the "pivot" is just a refinement.
+
+## Step 2 — Walk through Ries's 10 pivot types
+
+For each, ask: *"Does this match what your evidence is telling you?"* Cite brain entries as you go.
+
+1. **Zoom-in pivot** — A single feature becomes the whole product.
+   *Evidence pattern: users love ONE thing, ignore the rest.*
+2. **Zoom-out pivot** — The whole product becomes a single feature of something larger.
+   *Evidence pattern: users want it as part of an existing workflow, not standalone.*
+3. **Customer segment pivot** — Same product/problem, different customer.
+   *Evidence: wrong people responding well to right solution.*
+4. **Customer need pivot** — Same customer, different problem.
+   *Evidence: customer engaged but not on the pain you targeted.*
+5. **Platform pivot** — Application ↔ platform.
+   *Evidence: developers want APIs, or customers want a turnkey app.*
+6. **Business architecture pivot** — B2B ↔ B2C, or high-margin/low-volume ↔ low-margin/high-volume.
+   *Evidence: monetization model mismatch with how the segment buys.*
+7. **Value capture pivot** — Different monetization model.
+   *Evidence: willingness-to-pay exists but not for the model you priced.*
+8. **Engine of growth pivot** — Sticky / viral / paid (Ries's three engines).
+   *Evidence: growth working through a different mechanism than designed for.*
+9. **Channel pivot** — Same product, different distribution.
+   *Evidence: customers prefer to buy/discover differently than you offer.*
+10. **Technology pivot** — Same customer/problem, different tech.
+    *Evidence: you can deliver the same value cheaper/faster/better with new tech.*
+
+## Step 3 — Kill criteria for the OLD hypothesis (non-negotiable)
+
+Before any pivot, force the founder to write down:
+- What was the original hypothesis?
+- What evidence proved it wrong (cite specific interview / metrics)?
+- Why are we sure it's wrong, not just early?
+
+Pivots without explicit kill criteria are noise. **Refuse to log the pivot until these are written down.**
+
+## Step 4 — New hypothesis with falsification
+
+The pivot is a NEW hypothesis. It needs:
+- Specific customer segment
+- Specific pain
+- Falsification criteria (what would prove THIS wrong)
+
+Run `/problem` on the new hypothesis. **Don't auto-overwrite the old one** — preserve history. Old `problem.md` becomes part of the pivot log.
+
+## Step 5 — Log to brain/pivots.md
+
+Append:
+
+```markdown
+## <date> — <pivot type>: <one-liner>
+
+### Old hypothesis (killed)
+<copy of brain/problem.md before pivot>
+
+### Kill criteria met
+- <specific evidence — interview ref, metric, quote>
+- <...>
+
+### Pivot type (Ries)
+<one of 10>
+
+### New hypothesis
+<see brain/problem.md as of this date>
+
+### What carries over
+- Believers willing to follow: <names>
+- Evidence still relevant: <ref>
+
+### What we lose
+- <ref>
+
+### Reasoning
+<one paragraph — what we learned, what we believe now, what we'll watch>
+```
+
+## Anti-pattern to call out
+
+If founder wants to pivot every 2 weeks: this is thrash, not pivoting. Surface `brain/pivots.md` history. *"You've pivoted twice in 8 weeks. The problem is probably not the hypothesis — it's that you're not running long enough experiments to learn from any of them. Either let this one run another 4 weeks, or admit it's not the hypothesis you're rejecting — it's the discomfort of a hypothesis without quick wins."*
+
+## Why this exists
+
+Eric Ries: *"A pivot is a structured course correction designed to test a new fundamental hypothesis about the product, strategy, and engine of growth."*
+
+Most founders pivot in panic. This skill makes pivots **structured, evidence-driven, and historical** — the brain remembers every old hypothesis and why it was killed. Future-you reads pivots.md and sees the trail of validated learning, not just the latest narrative.
+
+Uri Levine: *"It's only a pivot if you can articulate what was wrong. Otherwise it's a fresh start, which is fine — just call it that."*

.claude/skills/pmf-check/SKILL.md

@@ -0,0 +1,138 @@
+---
+name: pmf-check
+description: Triangulate Product-Market Fit signal from three independent sources — Sean Ellis test, retention cohorts, believer density. Returns Strong PMF / PMF / Pre-PMF / PMF Mirage. The triangulation is what's hard to fake.
+---
+
+# /pmf-check
+
+Read `brain/believers/index.md`, `brain/interviews/`, `brain/metrics/` (if exists), `brain/runway.md`. Use `believer_density()` from the brain MCP.
+
+PMF is **not one number**. It's three signals that have to align. Any single signal alone is gameable. The triangulation is what's hard to fake.
+
+## Three signals
+
+### Signal 1 — Sean Ellis test (the survey)
+
+Ask **all current users** (not cherry-picked):
+> *"How would you feel if you could no longer use this product?"*
+
+Options:
+- Very disappointed
+- Somewhat disappointed
+- Not disappointed
+- N/A — I no longer use it
+
+**Threshold: ≥40% "very disappointed" = PMF signal.**
+
+If founder hasn't run this:
+- Generate the survey form (Google Form / Typeform / Tally — the questions, not the platform)
+- Tell them to send to all active users (not just believers)
+- Aim for 30+ responses minimum
+
+If founder has results, capture in `brain/metrics/sean-ellis.md`:
+
+```markdown
+# Sean Ellis Test
+
+Captured: <date>
+Sample size: <N responses>
+Sent to: <user pool description — must be ALL users, not selected>
+
+## Results
+- Very disappointed: X% (N=...)
+- Somewhat disappointed: Y% (N=...)
+- Not disappointed: Z% (N=...)
+- N/A: ...% (N=...)
+
+## Verdict
+<PASS (≥40%) / WEAK (25-39%) / FAIL (<25%)>
+```
+
+### Signal 2 — Retention cohort
+
+Day 7, 30, 90 retention curves. **The shape matters more than the number.**
+
+- **Flat at zero** → no PMF. Users churn completely.
+- **Curve that flattens above zero** → PMF signal. Some core group keeps using it.
+- **Rising curve** → exceptional, rare. (Network effects working.)
+
+Calculate from product analytics. If founder doesn't have this set up:
+- Ask for raw data (CSV: `user_id, signup_date, last_active_date`, plus weekly active flags if available)
+- Compute cohort retention manually (Day 1, 7, 14, 30, 60, 90)
+- Output to `brain/metrics/retention-cohorts.md`
+
+Verdict: **PASS** (curve flattens >0%), **WEAK** (asymptote near zero), **FAIL** (everyone gone by Day 30).
+
+### Signal 3 — Believer density (from /believers)
+
+Run `/believers` if not recently run. Use `believer_density()` from brain MCP.
+
+- **<15%** → PMF mirage (most users are tire-kickers)
+- **15–30%** → weak; refine targeting
+- **30–50%** → healthy
+- **≥50%** → strong, focused niche
+
+Verdict: **PASS** (≥30%), **WEAK** (15-29%), **FAIL** (<15%).
+
+## Triangulation
+
+Mark each signal:
+
+```
+              | Threshold        | Actual   | Verdict
+--------------|------------------|----------|--------
+Sean Ellis    | ≥40%             | __%      | ____
+Retention     | curve flattens   | ____     | ____
+Believer dens | ≥30%             | __%      | ____
+```
+
+## Final verdicts
+
+- **3 of 3 PASS → STRONG PMF.** Stage 2 → 3 gate met. Time to scale. Run `/default-alive` and `/story`; raise if needed.
+- **2 of 3 PASS → PMF.** Real signal but specific weakness. Fix the FAIL before scaling.
+- **1 of 3 PASS → Pre-PMF.** Early. Keep iterating in Stage 2. Don't scale yet.
+- **0 PASS, or only believer density passes → PMF MIRAGE.** Most dangerous state. Feels like PMF but isn't. Vanity metric pattern. Re-target or refine hypothesis.
+
+## Output: brain/metrics/pmf-check.md
+
+```markdown
+# PMF check
+
+Captured: <date>
+
+## Sean Ellis
+<copy of brain/metrics/sean-ellis.md results>
+Verdict: ___
+
+## Retention
+<cohort table>
+Verdict: ___
+
+## Believer density
+<from believer_density()>
+Verdict: ___
+
+## Triangulation
+<2 of 3 / 3 of 3 / etc.>
+**Final: <Strong PMF / PMF / Pre-PMF / PMF Mirage>**
+
+## Action
+<one of: scale / address weakness X / iterate / re-target>
+```
+
+## Anti-patterns to call out
+
+- **"I have 1000 signups" alone is not PMF** — *"Users without retention is acquisition, not PMF."*
+- **Sean Ellis from cherry-picked users** — *"Must be ALL active users, not selected ones. Otherwise the bias makes it noise."*
+- **Founder claims PMF before this skill runs** — *"PMF is the triangulation, not the vibe. Run the three checks."*
+- **One signal screaming, two silent** — *"That's a vanity metric, not PMF. The triangulation is the safeguard."*
+
+## Why this exists
+
+- Eric Ries: *"Vanity metrics make you feel good, actionable metrics tell you the truth."*
+- Sean Ellis (the test's author): *"Until you hit 40%, scaling acquisition is multiplying churn."*
+- Uri Levine: *"Most founders think they have PMF when they have one happy customer per fifty. PMF is when most users would be heartbroken if you shut down — not when one user said something nice."*
+
+This skill is the reality check before the scaling decision.
+
+After `/pmf-check` returns Strong PMF: stage gate to 3 unlocks. Run `/default-alive` next to check whether you can fund the scale-up.