Agent: Smart Agent Detector

Agent ID: smart-agent-detector Priority: 100 Version: 1.0.0 Status: Active

---

Purpose

Advanced intelligent agent selection system using multi-layer analysis of natural language, context, file patterns, and project structure to automatically select appropriate agents.

---

Multi-Layer Detection System

Layer	Name	Description
1	Natural Language	Analyze user message for intent, domain, complexity
2	Context Analysis	CWD, recent files, project structure, conversation history
3	Domain Patterns	Match task patterns against agent capabilities
4	Confidence Scoring	Weighted multi-criteria algorithm
5	Recommendation	Primary, secondary, optional agents with reasoning

---

Scoring Algorithm

Weights

Criterion	Points	Description
Explicit Mention	+60	User directly mentions technology
Keyword Exact Match	+50	Direct keyword match to intent
Project Context	+40	CWD, file structure, package files
Semantic Match	+35	Contextual/implied match
Task Complexity	+30	Inferred complexity level
Conversation History	+25	Previous context, active agents
File Patterns	+20	Recent files, naming conventions

Thresholds

Role	Score	Description
Primary Agent	≥80	Lead development
Secondary Agent	50-79	Supporting role
Optional Agent	30-49	Nice-to-have
Not Activated	<30	Insufficient match

---

Intent Detection Patterns

Intent	Keywords	Primary Agent	Secondary
Implementation	implement, create, add, build, develop	Dev agent	qa-automation, ui-designer
Bug Fix	fix, bug, error, issue, broken, debug	Dev agent	qa-automation
Refactoring	refactor, improve, optimize, clean up	Dev agent	qa-automation
Testing	test, testing, QA, coverage, spec	qa-automation	Dev agent
Design/UI	design, UI, UX, figma, layout, styling	ui-designer	Dev agent
Documentation	document, docs, README, explain	pm-operations-orchestrator	-
Database	database, schema, migration, query, SQL	database-specialist	Backend agent
Security	security, vulnerability, audit, OWASP	security-expert	Dev agent
Performance	performance, slow, optimize, lighthouse	devops-cicd	Dev agent
Deployment	deploy, docker, kubernetes, CI/CD, pipeline	devops-cicd	-
Monitoring	monitor, errors, logs, sentry, tracking	devops-cicd	-

---

Technology Detection

Mobile Frameworks

Framework	Keywords	Files	Patterns
React Native	react-native, expo, RN	app.json (expo), package.json	*.phone.tsx, *.tablet.tsx
Flutter	flutter, dart, bloc, riverpod	pubspec.yaml	*.dart, lib/

Web Frameworks

Framework	Keywords	Files	Patterns
Angular	angular, ngrx, rxjs, signals	angular.json	*.component.ts, *.service.ts
Vue.js	vue, pinia, nuxt, composition api	*.vue files	composables/, stores/
React	react, hooks, context	package.json (react, no next)	*.jsx, *.tsx
Next.js	next, nextjs, ssr, ssg, app router	next.config.js	app/, pages/, route.ts

Backend Frameworks

Framework	Keywords	Files	Patterns
Node.js	nodejs, express, nestjs, fastify	package.json	*.controller.ts, *.module.ts
Python	python, django, fastapi, flask	requirements.txt, pyproject.toml	views.py, models.py
Go	go, golang, gin, fiber, grpc	go.mod, go.sum	*.go, handler.go
Laravel	laravel, php, eloquent, artisan	artisan, composer.json	*Controller.php, routes/

Specialized Agents

Agent	Keywords	Files/Patterns
database-specialist	database, schema, migration, SQL, postgres	migrations/, schema.sql
security-expert	security, vulnerability, OWASP, XSS, pentest	-
devops-cicd	docker, kubernetes, terraform, pipeline	Dockerfile, .github/workflows/
qa-automation	test, testing, coverage, jest, pytest	*.test.ts, __tests__/
ui-designer	design, UI, UX, figma, wireframe	design/, Figma URLs

---

QA Agent Conditional Activation

Activate When

Condition	Points
User explicitly requests testing	+50 (Primary)
Implementation intent + test infrastructure exists	+30 (Optional)
Bug fix intent + test infrastructure exists	+35 (Optional)

Skip When

• No test infrastructure detected (no test framework, config, or directory)
• User explicitly skips testing ("don't worry about tests")
• Documentation-only or deployment-only task

Test Infrastructure Detection

Language	Check For
JavaScript	jest.config.js, vitest.config.ts, cypress.config.js, package.json scripts.test
Python	pytest.ini, setup.cfg, requirements.txt (pytest)
PHP	phpunit.xml, composer.json (phpunit/phpunit)
Go	*_test.go files
General	tests/, __tests__/, test/, spec/ directories

---

Context Analysis

Project Detection Logic

1. Check package.json dependencies:
   - react-native → mobile-react-native (+40)
   - @angular/core → web-angular (+40)
   - vue → web-vuejs (+40)
   - react (no next) → web-reactjs (+40)
   - next → web-nextjs (+40)
   - express/nestjs → backend-nodejs (+40)

2. Check other package files:
   - pubspec.yaml → mobile-flutter (+40)
   - requirements.txt/pyproject.toml → backend-python (+40)
   - go.mod → backend-go (+40)
   - artisan → backend-laravel (+40)

3. Analyze recent files:
   - *.vue → web-vuejs (+20)
   - *.dart → mobile-flutter (+20)
   - *.go → backend-go (+20)
   - *.py → backend-python (+20)
   - *.test.*, *.spec.* → qa-automation (+20)

4. Conversation history:
   - Active agents → boost (+25)
   - Previous tech mentions → boost (+20)

---

Examples

Example 1: Implicit Mobile

User: "Add share button to post screen"

Analysis:
- "screen" → mobile terminology (+35)
- CWD: /mobile-app (+40)
- Recent: PostScreen.phone.tsx (+20)

Result:
✅ mobile-react-native: 95 pts (Primary)
✅ ui-designer: 50 pts (Secondary) - "button" UI
✅ qa-automation: 30 pts (Optional) - if test infra exists

Example 2: Full-Stack Feature

User: "Build user profile page with API"

Analysis:
- "page" → frontend (+35)
- "API" → backend (+50)
- CWD: mixed project

Result:
✅ web-reactjs: 55 pts (Primary - Frontend)
✅ backend-nodejs: 55 pts (Primary - Backend)
✅ ui-designer: 45 pts (Secondary)

Example 3: Vague Request

User: "Fix the login issue"

Analysis:
- "fix" → bug fix intent
- CWD: /backend-api (Laravel)
- Recent: AuthController.php (+20)

Result:
✅ backend-laravel: 90 pts (Primary)
✅ qa-automation: 35 pts (Optional) - bug validation
❌ ui-designer: skipped - backend context

Example 4: Security Audit

User: "Check if authentication is secure"

Analysis:
- "secure" → security intent (+50)
- "authentication" → auth context (+35)

Result:
✅ security-expert: 85 pts (Primary)
✅ backend-nodejs: 45 pts (Secondary)

Example 5: No Test Infrastructure

User: "Add new landing page"

Analysis:
- "page" → web (+35)
- CWD: /website
- Test Infrastructure: ❌ None detected

Result:
✅ web-nextjs: 85 pts (Primary)
✅ ui-designer: 50 pts (Secondary)
❌ qa-automation: 0 pts (SKIPPED - no test infra)

Example 6: Explicit Test Skip

User: "Add feature, don't worry about tests"

Analysis:
- "feature" → implementation
- User explicitly skips testing

Result:
✅ web-reactjs: 75 pts (Primary)
❌ qa-automation: 0 pts (SKIPPED - user request)

---

Always Active Agents

These agents don't require scoring:

• pm-operations-orchestrator - Workflow coordination
• project-detector - Auto-detect project type
• project-context-manager - Context tracking

---

Output Format

🧠 **Agent Selection Complete**

**Primary Agents:**
- 🤖 [agent-name] (Score: XX) - [role]
  Reasoning: [why selected]

**Secondary Agents:**
- 🎨 [agent-name] (Score: XX) - [role]

**Optional Agents:**
- ✅ [agent-name] (Score: XX) - [role]

**Confidence:** [High/Medium/Low] (XX%)

---

Fallback Strategies

Scenario	Action
No agents ≥30 pts	Ask user for clarification
Multiple agents tie	Prefer higher base priority
Context unclear	Show top 3 options, let user choose

---

Related Documentation

• Skill: skills/agent-detector/agent-selection.md
• Guide: docs/AGENT_SELECTION_GUIDE.md
• Agent Catalog: README.md

---

Version: 1.0.0 | Last Updated: 2025-11-28