$ ctx leaderboard --hallucination
Hallucination Leaderboard
Repos: 12
Tasks: 20

System              Correct Skill
------------------  -------------
Raw Agent           10.0%
BackendGuard + Codex   80.0%

Sample failures:
- expo-eas: "fix deployed"
  expected: eas, mobile-deployment, github-actions-ci-cd
  raw (wrong): eas, env-secret-management, railway-render-deployment
  backendguard (correct): eas, github-actions-ci-cd, mobile-deployment
- next-vercel: "fix deployed"
  expected: vercel-deployment, github-actions-ci-cd, env-secret-management
  raw (wrong): eas, env-secret-management, railway-render-deployment
  backendguard (correct): vercel-deployment, github-actions-ci-cd, env-secret-management
- docker-node: "docker image build failed"
  expected: docker, build-log-debugging
  raw (wrong): docker, build-log-debugging, github-actions-ci-cd
  backendguard (correct): build-log-debugging, docker
- railway-render: "Railway deploy health check failed"
  expected: railway-render-deployment, build-log-debugging
  raw (wrong): railway-render-deployment, build-log-debugging, firebase-hosting
  backendguard (correct): build-log-debugging, railway-render-deployment
- firebase-hosting: "deploy firebase hosting"
  expected: firebase-hosting
  raw (wrong): firebase-hosting, flutter-firebase, railway-render-deployment
  backendguard (correct): firebase-hosting
- nest-prisma: "optimize slow prisma queries"
  expected: prisma, nestjs-module
  raw (wrong): prisma, nestjs-module, android-signing
  backendguard (correct): nestjs-module, prisma

$ ctx skills doctor -- "fix deployed"  # Expo fixture
BackendGuard skill doctor
fixture: expo-eas
prompt: fix deployed
selected: eas, github-actions-ci-cd, mobile-deployment
expected: eas, mobile-deployment, github-actions-ci-cd
rejected: vercel-deployment

$ ctx skills doctor -- "fix deployed"  # Next/Vercel fixture
BackendGuard skill doctor
fixture: next-vercel
prompt: fix deployed
selected: vercel-deployment, github-actions-ci-cd, env-secret-management
expected: vercel-deployment, github-actions-ci-cd, env-secret-management
rejected: eas
