Replace ScaffBench with Fixproof, a sealed benchmark graded by hidden tests

.github/workflows/e2e-test.yaml

@@ -22,10 +22,6 @@ on:
         description: "Run deep smoke tests for the AI search benchmark template"
         type: boolean
         default: true
-      run_solvability:
-        description: "Run ScaffBench per-spec solvability gate (all ecosystems)"
-        type: boolean
-        default: true
 
 permissions:
   contents: read
@@ -399,69 +395,6 @@ jobs:
           if-no-files-found: warn
           retention-days: 7
 
-  scaffbench-solvability:
-    name: ScaffBench Spec Solvability
-    if: ${{ github.event_name == 'schedule' || (github.event_name == 'workflow_dispatch' && inputs.run_solvability) }}
-    runs-on: ubuntu-latest
-    timeout-minutes: 90
-    steps:
-      - uses: actions/checkout@v4
-
-      - uses: oven-sh/setup-bun@v2
-        with:
-          bun-version: latest
-
-      - uses: actions/setup-node@v4
-        with:
-          node-version: "22"
-
-      - uses: astral-sh/setup-uv@v5
-
-      - uses: erlef/setup-beam@v1
-        with:
-          otp-version: "27"
-          elixir-version: "1.18"
-
-      - uses: actions/setup-dotnet@v4
-        with:
-          dotnet-version: "10.0.x"
-
-      - name: Install protoc
-        # The rust-leptos-axum spec uses Tonic/gRPC, whose prost build script
-        # compiles .proto files with protoc at `cargo check` time.
-        run: sudo apt-get update && sudo apt-get install -y protobuf-compiler
-
-      - name: Show toolchains
-        run: |
-          echo "cargo:  $(cargo --version 2>&1 || echo MISSING)"
-          echo "go:     $(go version 2>&1 || echo MISSING)"
-          echo "uv:     $(uv --version 2>&1 || echo MISSING)"
-          echo "dotnet: $(dotnet --version 2>&1 || echo MISSING)"
-          echo "elixir: $(elixir --version 2>&1 || echo MISSING)"
-          echo "mix:    $(mix --version 2>&1 || echo MISSING)"
-          echo "protoc: $(protoc --version 2>&1 || echo MISSING)"
-
-      - uses: actions/cache@v4
-        with:
-          path: |
-            node_modules
-            ~/.bun/install/cache
-          key: solvability-${{ runner.os }}-bun-${{ hashFiles('bun.lock') }}
-          restore-keys: solvability-${{ runner.os }}-bun-
-
-      - name: Install Dependencies
-        run: bun install --frozen-lockfile
-
-      - name: Build Packages
-        run: |
-          bun run --cwd packages/types build
-          bun run --cwd packages/project-lifecycle build
-          bun run --cwd packages/template-generator build
-          bun run --cwd apps/cli build
-
-      - name: Run Spec Solvability Gate
-        run: bun test apps/cli/test/e2e/scaffbench-solvability.test.ts --timeout 1200000
-
   playwright:
     name: Playwright Web Builder Tests
     if: ${{ github.event_name != 'workflow_dispatch' || inputs.run_playwright }}

.gitignore

@@ -52,7 +52,6 @@ yarn.lock
 .smoke-manifest.json
 testing/.release-guard/
 testing/.published-package/
-testing/.tmp-scaffbench-*/
 reports/
 artifacts/demos/
 COMPARISON-PLAN.md

AGENTS.md

@@ -27,7 +27,6 @@ See `docs/guidelines/` for deeper reference on these topics:
 - `template-output-and-validation.md` - template conditional logic, generated output validation, sync test discipline, and framework-specific constraints
 - `remotion-video-style.md` - default visual style, color system, motion rules, and branding for Remotion videos in this project
 - `design-reading-guide.md` - ordered index of design-related markdown (agent skills + BF video style), precedence when sources conflict, and commands to verify coverage
-- `scaffbench-benchmark.md` - ScaffBench protocol, execution, validation, and publication rules
 - `adding-new-tool-options/` - **read this subfolder when adding any new library, tool, or category** to any ecosystem (TypeScript, Rust, Go, Python). Covers every file that must be touched, with worked examples, template handler reference, test patterns, and routing edge cases (Convex skips, self-backend, frontend array detection, processor ordering)
 
 ## Web UI

README.md

@@ -70,7 +70,7 @@ Connect your AI coding agents to Better Fullstack with one command:
 npx create-better-fullstack@latest install
 ```
 
-After install, supported agents scaffold projects through the MCP server, 2.6x faster than hand-writing in ScaffBench; see the [AI docs](https://better-fullstack.dev/docs/ai/overview) for setup.
+After install, supported agents scaffold projects through the MCP server; see the [AI docs](https://better-fullstack.dev/docs/ai/overview) for setup.
 
 <br>
 

apps/cli/tsconfig.json

@@ -12,7 +12,6 @@
     "types": ["node"],
     "paths": {
       "@/*": ["./src/*"],
-      "@scaffbench/*": ["../../scripts/scaffbench/*"],
       "@test/*": ["./test/*"],
       "@testing/*": ["../../testing/*"],
       "@web/*": ["../web/src/*"]

apps/web/content/docs/verification.mdx

@@ -27,7 +27,7 @@ The builder and stack pages read the option-level projection from
 recipe remain listed and show their limitation. A template, schema, dependency, toolchain, or
 deployed-commit change needs a new receipt.
 
-## Separate from ScaffBench
+## Separate from Fixproof
 
-ScaffBench evaluates how coding models respond to scaffold tasks. It does not verify a Better
+Fixproof evaluates coding agents on sealed issues with hidden tests. It does not verify a Better
 Fullstack release, and it does not raise the evidence level shown here.