AI-consumed reference. Optimized for Claude to read during execution. Human-readable explanation: see docs/architecture/HIERARCHICAL_PLANNING.md or docs/getting-started/ depending on topic.
Lazy Agent Loader
Reduce token usage: load agent summaries initially (~50 tokens each), full definition only when activated.
Agent Index
agent_index[16]{id,category,specialty,keywords}:
mobile,dev,React Native/Expo mobile,react-native/expo/RN/mobile/ios/android
mobile-flutter,dev,Flutter/Dart mobile,flutter/dart/bloc/mobile
web-angular,dev,Angular frontend,angular/ngrx/rxjs/typescript
web-vuejs,dev,Vue.js frontend,vue/vuejs/pinia/nuxt/composition
web-reactjs,dev,React frontend,react/reactjs/jsx/hooks/redux
web-nextjs,dev,Next.js fullstack,next/nextjs/ssr/ssg/app-router
architect,dev,Node.js backend,nodejs/express/nestjs/fastify/api
backend-python,dev,Python backend,python/django/fastapi/flask/api
backend-go,dev,Go backend,go/golang/gin/fiber/api
backend-laravel,dev,Laravel/PHP backend,laravel/php/eloquent/artisan
security,quality,Security auditing,security/vulnerability/audit/owasp
tester,quality,Testing/QA,test/testing/coverage/qa/jest/cypress
devops,ops,DevOps/CI-CD,deploy/docker/kubernetes/ci-cd/pipeline
router,infra,Agent detection,detect/agent/select/route
lead,infra,PM/orchestration,pm/project/orchestrate/manage
scanner,infra,Project management,project/detect/identify/config/context
Loading Strategy
| Score | Level | Action |
|-------|-------|--------|
| >= 80 | PRIMARY | Load full definition (agents/[id].md) |
| 50-79 | SECONDARY | Summary only from index |
| < 50 | OPTIONAL | Don't load |
Token Savings
comparison[4]{scenario,without_lazy,with_lazy,savings}:
Initial load,~48000,~1200,97.5%
Single agent,~48000,~2700,94.4%
Dual agent,~48000,~4200,91.3%
3 agents,~48000,~5700,88.1%
Cache
Loaded agents cached in session. Track: loaded_agents[]: mobile,tester. Force reload: reload agent <id>.
Integration
Used automatically by agent-detector for optimized loading. Score agents by keywords from index, load only PRIMARY agents.