Rename to Build Gate and sharpen proof gate

Author: dicnunz

README.md

@@ -1,37 +1,51 @@
-# Codex Venture Surgeon Skill
+# Codex Build Gate Skill
 
-`venture-surgeon` is a Codex skill for deciding whether a startup or product idea deserves build time.
+`build-gate` is a Codex skill that stops Codex from building a startup idea before the idea clears a proof gate.
 
-It is built for evidence-first triage: buyer clarity, current behavior, first customers, competition, founder edge, a 2-week MVP contract, and explicit kill criteria.
+It is built for the moment right before implementation: buyer clarity, current behavior, proof gaps, first buyers, a two-week validation contract, and a hard no-build clause.
 
 ## Why this exists
 
-Most startup validators produce confident theater: big-market language, generic scorecards, and no hard stop line.
+Most startup validators produce confident theater: big-market language, generic scorecards, and a verdict that still lets the agent go build the wrong thing.
 
-This skill is stricter. It forces the agent to separate provided facts, verified facts, assumptions, and unknowns before recommending build/validate/pivot/kill.
+Build Gate is stricter. It makes the agent say what proof is missing and what must happen before build time is allowed.
+
+## Why use this one
+
+Simple pressure-test skills are useful if you want a harsh opinion.
+
+Build Gate is for when you want an operating constraint:
+
+- `Pass / Hold / Pivot / Kill`, not just "strong" or "weak".
+- Proof ledger with `Provided`, `Verified`, `Assumed`, and `Unknown` claims.
+- No-build clause that blocks implementation until the riskiest proof exists.
+- Current-behavior map, because the real competitor is often a spreadsheet, agency, internal process, or doing nothing.
+- Next-10-buyer moves instead of vague "talk to users" advice.
+- Two-week validation contract with pass/fail/kill criteria.
+- Non-destructive installer that refuses to overwrite an existing skill unless `--force` creates a backup.
 
 ## What it does
 
-- Gives a direct `Build`, `Validate first`, `Pivot`, or `Kill` decision.
-- Builds an evidence ledger instead of pretending weak signals are proof.
-- Scores pain, buyer clarity, urgency, current workaround, differentiation, distribution, validation speed, and founder edge.
+- Gives a direct `Pass`, `Hold`, `Pivot`, or `Kill` gate result.
+- Builds a proof ledger instead of pretending weak signals are proof.
+- Scores pain, buyer clarity, urgency, current workaround, differentiation, reachability, validation speed, and founder edge.
 - Maps current behavior as the real competition.
-- Designs a first-10-customer route before ads or launch theater.
-- Defines a 2-week MVP contract with build/cut/pass/fail lines.
+- Designs next-10-buyer moves before ads or launch theater.
+- Defines a two-week validation contract with test/cut/pass/fail lines.
 - Browses or asks for source proof when current market facts matter.
 
 ## Install from GitHub
 
 ```bash
-npx --yes github:dicnunz/codex-venture-surgeon-skill
+npx --yes github:dicnunz/codex-build-gate-skill
 ```
 
 Then restart Codex so it discovers the new skill.
 
 Use:
 
 ```text
-Use $venture-surgeon to decide whether this deserves build time:
+Use $build-gate before building this:
 
 Idea: AI bookkeeping cleanup for solo founders.
 Customer: US solo founders using QuickBooks.
@@ -40,56 +54,56 @@ Desired behavior: pay $49 for a one-time cleanup.
 
 ## Safer installer behavior
 
-The installer copies `venture-surgeon/` into:
+The installer copies `build-gate/` into:
 
 ```text
-~/.codex/skills/venture-surgeon
+~/.codex/skills/build-gate
 ```
 
 It refuses to overwrite an existing install unless you pass `--force`. With `--force`, it first renames the old folder to a timestamped backup.
 
 ```bash
-npx --yes github:dicnunz/codex-venture-surgeon-skill --force
+npx --yes github:dicnunz/codex-build-gate-skill --force
 ```
 
 Preview without writing:
 
 ```bash
-npx --yes github:dicnunz/codex-venture-surgeon-skill --dry-run
+npx --yes github:dicnunz/codex-build-gate-skill --dry-run
 ```
 
 Install somewhere else:
 
 ```bash
-npx --yes github:dicnunz/codex-venture-surgeon-skill --skills-dir ./skills
+npx --yes github:dicnunz/codex-build-gate-skill --skills-dir ./skills
 ```
 
 ## Manual install
 
 ```bash
-git clone https://github.com/dicnunz/codex-venture-surgeon-skill.git
+git clone https://github.com/dicnunz/codex-build-gate-skill.git
 mkdir -p ~/.codex/skills
-cp -R codex-venture-surgeon-skill/venture-surgeon ~/.codex/skills/venture-surgeon
+cp -R codex-build-gate-skill/build-gate ~/.codex/skills/build-gate
 ```
 
 Restart Codex.
 
 ## Modes
 
-- `triage`: fast build/validate/pivot/kill decision.
-- `evidence-map`: assumptions, proof, disconfirming signals.
-- `competition`: current behavior, competitors, substitutes, switching cost.
-- `customer-route`: first 10 customers and manual conversion route.
-- `mvp-contract`: 2-week MVP with explicit cuts and pass/fail.
-- `positioning`: wedge, category, buyer language, credibility proof.
+- `gate`: fast Pass/Hold/Pivot/Kill decision.
+- `proof`: assumptions, proof, disconfirming signals.
+- `behavior`: current behavior, competitors, substitutes, switching cost.
+- `first-buyers`: next 10 reachable buyers and manual conversion route.
+- `validation-contract`: two-week validation plan with explicit cuts and pass/fail.
+- `wedge`: switching reason, buyer language, category, credibility proof.
 - `full`: compact all-in-one diagnosis.
 
 ## Validation packet helper
 
 Create a reusable Markdown packet:
 
 ```bash
-node venture-surgeon/scripts/make-validation-packet.js \
+node build-gate/scripts/make-validation-packet.js \
   --idea "AI bookkeeping cleanup for solo founders" \
   --customer "US solo founders using QuickBooks" \
   --outcome "pay $49 for a one-time cleanup" \
@@ -101,7 +115,7 @@ node venture-surgeon/scripts/make-validation-packet.js \
 ```bash
 npm test
 npm run pack:check
-python3 /path/to/quick_validate.py venture-surgeon
+python3 /path/to/quick_validate.py build-gate
 ```
 
 ## Security notes

build-gate/SKILL.md

@@ -0,0 +1,124 @@
+---
+name: build-gate
+description: Pre-build evidence gate for startup and product ideas. Use when Codex is asked whether to build an idea, stop before implementation, require proof before coding, validate demand, inspect current customer behavior, choose first-customer moves, define a two-week validation contract, set kill criteria, or decide Pass/Hold/Pivot/Kill before writing product code.
+---
+
+# Build Gate
+
+Use this skill as a gate before Codex starts building. The job is to decide whether implementation is allowed yet, what proof is missing, and what smallest customer-facing move can open or close the gate.
+
+## First Move
+
+If the idea, target buyer/user, or desired behavior/payment is missing, ask one short question:
+
+```text
+Send the idea, target customer, and what they should do or pay for.
+```
+
+If those are present, start immediately.
+
+## Gate Modes
+
+Choose the mode from the user's wording:
+
+- `gate`: fast Pass/Hold/Pivot/Kill decision.
+- `proof`: claim ledger, proof needed, disconfirming evidence.
+- `behavior`: what customers do today, including manual workarounds and substitutes.
+- `first-buyers`: where to find 10 reachable buyers and what to ask them for.
+- `validation-contract`: a two-week test with build/cut/pass/fail lines.
+- `wedge`: buyer language, switching reason, category, credibility proof.
+- `full`: compact all-in-one gate review. Use this when the user asks generally or wants a serious pre-build review.
+
+For `full`, `deep`, unfamiliar markets, or any market/pricing/competitor claim that matters, read `references/operating-manual.md`.
+
+## Proof Standard
+
+Separate what is known from what is guessed:
+
+- `Provided`: facts from the user.
+- `Verified`: current facts checked from sources or live surfaces.
+- `Assumed`: plausible but unverified claims.
+- `Unknown`: facts that could flip the verdict.
+
+Browse or inspect live sources when current competitors, pricing, platform rules, funding, regulations, or recent market changes materially affect the decision. Do not invent market size, revenue, customers, endorsements, benchmark results, or private access.
+
+Default stance: the gate is closed until real behavior justifies build time.
+
+## Default Output
+
+Keep the default compact. Use this shape unless the user asks for more detail:
+
+```markdown
+**Gate**
+Pass / Hold / Pivot / Kill
+
+One sentence on whether Codex should build now.
+
+**Why**
+2 bullets, tied to evidence.
+
+**Proof Ledger**
+| Claim | Status | Why It Matters | Proof Needed |
+|---|---|---|---|
+| ... | Provided/Verified/Assumed/Unknown | ... | ... |
+
+**Scorecard**
+| Area | Score | Read |
+|---|---:|---|
+| Pain intensity | 1-5 | ... |
+| Buyer clarity | 1-5 | ... |
+| Urgency | 1-5 | ... |
+| Current workaround | 1-5 | ... |
+| Differentiation | 1-5 | ... |
+| Reachability | 1-5 | ... |
+| Validation speed | 1-5 | ... |
+| Founder edge | 1-5 | ... |
+
+**Locked Door**
+One sentence.
+
+**No-Build Clause**
+Do not implement beyond [scope] until [proof].
+
+**Next 10 Buyer Moves**
+1. ...
+2. ...
+3. ...
+
+**Two-Week Contract**
+- Test:
+- Manual/concierge:
+- Cut:
+- Pass:
+- Fail:
+```
+
+## Decision Rules
+
+- `Pass`: implementation is allowed because buyer, pain, current behavior, and a reachable path are credible.
+- `Hold`: the idea might work, but one proof item must be collected before serious build time.
+- `Pivot`: the nearby problem is real, but the stated buyer, wedge, or behavior is wrong.
+- `Kill`: no urgent buyer, no reachable early adopter, no credible switch reason, or no falsifiable two-week path.
+
+## Operating Rules
+
+- Be specific to the idea. No generic startup advice.
+- Treat current behavior as the real competition.
+- Treat compliments, waitlists, and "I'd use this" as weak evidence.
+- Prefer paid commitments, repeated manual pain, switching behavior, or real workflow access.
+- Keep first customers manual before ads, automation, or growth tricks.
+- Cut features that do not test the riskiest assumption.
+- Include exact kill criteria. A test without a kill line is theater.
+- If the idea is weak, say so and name the closest viable pivot.
+- If the user asks Codex to build after a `Hold`, define the smallest fake/manual surface that can gather proof without becoming the product.
+
+## Optional Packet Script
+
+Use `scripts/make-validation-packet.js` when the user wants a reusable Markdown packet for an idea:
+
+```bash
+node build-gate/scripts/make-validation-packet.js \
+  --idea "AI bookkeeping cleanup for solo founders" \
+  --customer "US solo founders using QuickBooks" \
+  --outcome "pay $49 for a one-time cleanup"
+```

build-gate/agents/openai.yaml

@@ -0,0 +1,4 @@
+interface:
+  display_name: "Build Gate"
+  short_description: "Pre-build evidence gate for startup ideas"
+  default_prompt: "Use $build-gate to decide whether this startup idea deserves build time."

…e-surgeon/references/operating-manual.md build-gate/references/operating-manual.mdventure-surgeon/references/operating-manual.md renamed to build-gate/references/operating-manual.md venture-surgeon/references/operating-manual.md renamed to build-gate/references/operating-manual.md

@@ -1,19 +1,19 @@
-# Venture Surgeon Operating Manual
+# Build Gate Operating Manual
 
-Load this file for full reports, unfamiliar markets, live competitor checks, or when a verdict depends on current facts.
+Load this file for full gate reviews, unfamiliar markets, live competitor checks, or when the build/no-build decision depends on current facts.
 
 ## Core Frame
 
-A startup idea earns build time only when four things line up:
+A startup idea passes the build gate only when four things line up:
 
 1. A specific person has a painful repeated problem.
 2. That person already spends money, time, status, or attention trying to solve it.
 3. The founder can reach enough of those people manually.
 4. A 2-week test can create evidence that changes the decision.
 
-If any one of those is absent, the output should shift from build planning to validation, pivoting, or killing.
+If any one of those is absent, keep the gate closed and prescribe the smallest proof move.
 
-## Evidence Quality Ladder
+## Proof Quality Ladder
 
 Rank evidence from strongest to weakest:
 
@@ -25,7 +25,7 @@ Rank evidence from strongest to weakest:
 6. Customer joins a waitlist or says they would use it.
 7. Founder thinks the market is large.
 
-Use weak evidence only as a reason to run a test, not as proof to build.
+Use weak evidence only as a reason to run a test, not as permission to build.
 
 ## Scorecard Guidance
 
@@ -38,19 +38,19 @@ Use 1-5 scores tied to evidence:
 | Urgency | No deadline | Seasonal or occasional | Immediate trigger or penalty |
 | Current workaround | None | Manual workaround | Paid/currently tolerated bad tool |
 | Differentiation | Feature clone | Sharper workflow | Clear wedge or unfair access |
-| Distribution access | Cold generic audience | Known channels | Direct access to early adopters |
+| Reachability | Cold generic audience | Known channels | Direct access to early adopters |
 | Validation speed | Months | Several weeks | Testable inside 14 days |
 | Founder edge | None | Some domain taste | Real access, proof, skill, or credibility |
 
 Do not average scores mechanically. One fatal zero can dominate the verdict.
 
 ## Mode Playbooks
 
-### triage
+### gate
 
-Output the decision, 3 strongest reasons, riskiest assumption, fastest test, and kill line.
+Output the gate result, 2 strongest reasons, locked door, no-build clause, fastest proof move, and kill line.
 
-### evidence-map
+### proof
 
 Build a table:
 
@@ -59,7 +59,7 @@ Build a table:
 
 Prefer tests that ask for money, workflow access, or behavior change.
 
-### competition
+### behavior
 
 Map:
 
@@ -71,7 +71,7 @@ Map:
 
 Browse when competitor existence, pricing, packaging, or recency matters.
 
-### customer-route
+### first-buyers
 
 Define:
 
@@ -83,7 +83,7 @@ Define:
 
 Avoid ads, launch platforms, generic communities, and growth hacks until manual reach works.
 
-### mvp-contract
+### validation-contract
 
 State:
 
@@ -96,7 +96,7 @@ State:
 
 Default MVP window is 14 days.
 
-### positioning
+### wedge
 
 Find the wedge by completing:
 
@@ -111,15 +111,15 @@ Keep positioning grounded in proof. Do not create grand category language before
 
 Use all of the following, compactly:
 
-1. Decision.
-2. Evidence ledger.
+1. Gate result.
+2. Proof ledger.
 3. Scorecard.
-4. Riskiest assumption.
-5. Failure modes.
-6. Competition/current behavior.
-7. First 10 customer route.
-8. MVP contract.
-9. Kill criteria.
+4. Locked door.
+5. No-build clause.
+6. Current behavior and substitutes.
+7. Next 10 buyer moves.
+8. Two-week contract.
+9. Kill criteria and pivot trigger.
 
 ## Kill Criteria Examples