intelligent-routing
Skill9 starsChanged 3 months agoApache-2.0View on GitHub
Automatic agent selection and intelligent task routing. Analyzes user requests and automatically selects the best specialist agent(s) without requiring explicit user mentions. Always active. Automatically selects the best specialist agent for each user request without explicit user mentions.
5 other projects use a version of this, 5 of them edited.
Adopted by
Most recently taken up by
- sid-lakhani/agents-configedited, first seen Oct 20, 2026
- vudovn/ag-kitedited, first seen Oct 8, 2026
- vudovn/antigravity-kitedited, first seen Oct 8, 2026
- jornalistainclusivo/jinc-skillsedited, first seen Apr 22, 2026
- SKB3002/agenthubedited, first seen Apr 20, 2026
Used with
Projects with it also use design-spec, verify-changes and 1 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
jornalistainclusivoSkilldesign-spec
How to author a DESIGN.md file — the machine-readable design-token + human-rationale format that must exist before any UI is built. YAML front-matter token schema (colors, typography, spacing, rounded, components), type system, token references, and canonical section order.
Also in 3 projects that use this
zzfancitizenSkillverify-changes
Verify code changes do not break the project by running tests, linter checks, and import validation. Use after making any code changes, editing Python files, modifying imports, updating dependencies, or refactoring.
Also in 3 projects that use this
OpenAISkillbrainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
Also in 3 projects that use this
Other versions5 projects use a version of this.
Word for word · 0
None found.
Edited · 5
- vudovn/ag-kit★ 8.2k95% same
- vudovn/antigravity-kit★ 8.2k95% same
- SKB3002/agenthub★ 496% same
- sid-lakhani/agents-config★ 2100% same
- jornalistainclusivo/jinc-skills★ 1100% same
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository36 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- api-designSkill
- architectureSkill
- bash-linuxSkill
- batch-operationsSkill
- behavioral-modesSkill
- claude-code-worktreesSkill
- clean-codeSkill
- context-compressionSkill
- coordinator-modeSkill
- database-designSkill
- deployment-proceduresSkill
- documentation-templatesSkill
- error-handlingSkill
- github-issuesSkill
- github-pmSkill
- github-projectSkill
- git-workflowSkill
- goal-builderSkill
- graphifySkill
- lint-and-validateSkill
- lint-fixSkill
- memory-systemSkill
- performance-profilingSkill
- powershell-windowsSkill
- pre-commit-verifySkill
- red-team-tacticsSkill
- release-vsixSkill
- run-testsSkill
- server-managementSkill
- setup-testingSkill
- testing-patternsSkill
- test-triageSkill
- uncle-bob-craftSkill
- vulnerability-scannerSkill
The file
We split this skill into the rules it gives an agent
8 rules. 6 are common, 1 is distinctive, and 1 is very specific to this project.
6
Common
Most of the 5 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Common: line 20
Before responding to ANY user request, perform automatic analysis:
- Common: line 59
Before responding to ANY request:
- Very specific: line 120
If request matches 2+ domains from different categories, automatically use orchestrator:
- Common: line 234
If task is unclear, still ask questions first
- Common: line 240
If GEMINI.md specifies explicit routing, follow it
- Common: line 290
No additional API calls required
- Common: line 297
If this is the first interaction in a project:
- Distinctive: line 309
Add to GEMINI.md temporarily:
“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!-- Vendored from ag-kit (github.com/vudovn/ag-kit) @ 20a13da6d4414c7c6ae33db050a9c606eaef9f40 :: .agents/skills/intelligent-routing/SKILL.md. MIT (c) vudovn. -->Intelligent Agent Routing
Purpose: Automatically analyze user requests and route them to the most appropriate specialist agent(s) without requiring explicit user mentions.
Core Principle
The AI should act as an intelligent Project Manager, analyzing each request and automatically selecting the best specialist(s) for the job.
How It Works
1. Request Analysis
Before responding to ANY user request, perform automatic analysis:
graph TD
A[User Request: Add login] --> B[ANALYZE]
B --> C[Keywords]
B --> D[Domains]
B --> E[Complexity]
C --> F[SELECT AGENT]
D --> F
E --> F
F --> G[security-auditor + backend-specialist]
G --> H[AUTO-INVOKE with context]2. Agent Selection Matrix
Use this matrix to automatically select agents:
| User Intent | Keywords | Selected Agent(s) | Auto-invoke? |
| Authentication | "login", "auth", "signup", "password" | security-auditor + backend-specialist | ✅ YES |
| UI Component | "button", "card", "layout", "style" | frontend-specialist | ✅ YES |
| Mobile UI | "screen", "navigation", "touch", "gesture" | mobile-developer | ✅ YES |
| API Endpoint | "endpoint", "route", "API", "POST", "GET" | backend-specialist | ✅ YES |
| Database | "schema", "migration", "query", "table" | database-architect + backend-specialist | ✅ YES |
| Bug Fix | "error", "bug", "not working", "broken" | debugger | ✅ YES |
| Test | "test", "coverage", "unit", "e2e" | test-engineer | ✅ YES |
| Deployment | "deploy", "production", "CI/CD", "docker" | devops-engineer | ✅ YES |
| Security Review | "security", "vulnerability", "exploit" | security-auditor + penetration-tester | ✅ YES |
| Performance | "slow", "optimize", "performance", "speed" | performance-optimizer | ✅ YES |
| Product Def | "requirements", "user story", "backlog", "MVP" | product-owner | ✅ YES |
| New Feature | "build", "create", "implement", "new app" | orchestrator → multi-agent | ⚠️ ASK FIRST |
| Complex Task | Multiple domains detected | orchestrator → multi-agent | ⚠️ ASK FIRST |
3. Automatic Routing Protocol
TIER 0 - Automatic Analysis (ALWAYS ACTIVE)
Before responding to ANY request:
// Pseudo-code for decision tree
function analyzeRequest(userMessage) {
// 1. Classify request type
const requestType = classifyRequest(userMessage);
// 2. Detect domains
const domains = detectDomains(userMessage);
// 3. Determine complexity
const complexity = assessComplexity(domains);
// 4. Select agent(s)
if (complexity === "SIMPLE" && domains.length === 1) {
return selectSingleAgent(domains[0]);
} else if (complexity === "MODERATE" && domains.length <= 2) {
return selectMultipleAgents(domains);
} else {
return "orchestrator"; // Complex task
}
}4. Response Format
When auto-selecting an agent, inform the user concisely:
🤖 **Applying knowledge of `@security-auditor` + `@backend-specialist`...**
[Proceed with specialized response]Benefits:
- ✅ User sees which expertise is being applied
- ✅ Transparent decision-making
- ✅ Still automatic (no /commands needed)
Domain Detection Rules
Single-Domain Tasks (Auto-invoke Single Agent)
| Domain | Patterns | Agent |
| Security | auth, login, jwt, password, hash, token | security-auditor |
| Frontend | component, react, vue, css, html, tailwind | frontend-specialist |
| Backend | api, server, express, fastapi, node | backend-specialist |
| Mobile | react native, flutter, ios, android, expo | mobile-developer |
| Database | prisma, sql, mongodb, schema, migration | database-architect |
| Testing | test, jest, vitest, playwright, cypress | test-engineer |
| DevOps | docker, kubernetes, ci/cd, pm2, nginx | devops-engineer |
| Debug | error, bug, crash, not working, issue | debugger |
| Performance | slow, lag, optimize, cache, performance | performance-optimizer |
| SEO | seo, meta, analytics, sitemap, robots | seo-specialist |
| Game | unity, godot, phaser, game, multiplayer | game-developer |
Multi-Domain Tasks (Auto-invoke Orchestrator)
If request matches 2+ domains from different categories, automatically use orchestrator:
Example: "Create a secure login system with dark mode UI"
→ Detected: Security + Frontend
→ Auto-invoke: orchestrator
→ Orchestrator will handle: security-auditor, frontend-specialist, test-engineerComplexity Assessment
SIMPLE (Direct agent invocation)
- Single file edit
- Clear, specific task
- One domain only
- Example: "Fix the login button style"
Action: Auto-invoke respective agent
MODERATE (2-3 agents)
- 2-3 files affected
- Clear requirements
- 2 domains max
- Example: "Add API endpoint for user profile"
Action: Auto-invoke relevant agents sequentially
COMPLEX (Orchestrator required)
- Multiple files/domains
- Architectural decisions needed
- Unclear requirements
- Example: "Build a social media app"
Action: Auto-invoke orchestrator → will ask Socratic questions
Implementation Rules
Rule 1: Silent Analysis
DO NOT announce "I'm analyzing your request..."
- ✅ Analyze silently
- ✅ Inform which agent is being applied
- ❌ Avoid verbose meta-commentary
Rule 2: Inform Agent Selection
DO inform which expertise is being applied:
🤖 **Applying knowledge of `@frontend-specialist`...**
I will create the component with the following characteristics:
[Continue with specialized response]Rule 3: Seamless Experience
The user should not notice a difference from talking to the right specialist directly.
Rule 4: Override Capability
User can still explicitly mention agents:
User: "Use @backend-specialist to review this"
→ Override auto-selection
→ Use explicitly mentioned agentEdge Cases
Case 1: Generic Question
User: "How does React work?"
→ Type: QUESTION
→ No agent needed
→ Respond directly with explanationCase 2: Extremely Vague Request
User: "Make it better"
→ Complexity: UNCLEAR
→ Action: Ask clarifying questions first
→ Then route to appropriate agentCase 3: Contradictory Patterns
User: "Add mobile support to the web app"
→ Conflict: mobile vs web
→ Action: Ask: "Do you want responsive web or native mobile app?"
→ Then route accordinglyIntegration with Existing Workflows
With /orchestrate Command
- User types
/orchestrate: Explicit orchestration mode - AI detects complex task: Auto-invoke orchestrator (same result)
Difference: User doesn't need to know the command exists.
With Socratic Gate
- Auto-routing does NOT bypass Socratic Gate
- If task is unclear, still ask questions first
- Then route to appropriate agent
With GEMINI.md Rules
- Priority: GEMINI.md rules > intelligent-routing
- If GEMINI.md specifies explicit routing, follow it
- Intelligent routing is the DEFAULT when no explicit rule exists
Testing the System
Test Cases
Test 1: Simple Frontend Task
User: "Create a dark mode toggle button"
Expected: Auto-invoke frontend-specialist
Verify: Response shows "Using @frontend-specialist"Test 2: Security Task
User: "Review the authentication flow for vulnerabilities"
Expected: Auto-invoke security-auditor
Verify: Security-focused analysisTest 3: Complex Multi-Domain
User: "Build a chat application with real-time notifications"
Expected: Auto-invoke orchestrator
Verify: Multiple agents coordinated (backend, frontend, test)Test 4: Bug Fix
User: "Login is not working, getting 401 error"
Expected: Auto-invoke debugger
Verify: Systematic debugging approachPerformance Considerations
Token Usage
- Analysis adds ~50-100 tokens per request
- Tradeoff: Better accuracy vs slight overhead
- Overall SAVES tokens by reducing back-and-forth
Response Time
- Analysis is instant (pattern matching)
- No additional API calls required
- Agent selection happens before first response
User Education
Optional: First-Time Explanation
If this is the first interaction in a project:
💡 **Tip**: I am configured with automatic specialist agent selection.
I will always choose the most suitable specialist for your task. You can
still mention agents explicitly with `@agent-name` if you prefer.Debugging Agent Selection
Enable Debug Mode (for development)
Add to GEMINI.md temporarily:
## DEBUG: Intelligent Routing
Show selection reasoning:
- Detected domains: [list]
- Selected agent: [name]
- Reasoning: [why]Summary
intelligent-routing skill enables:
✅ Zero-command operation (no need for /orchestrate)
✅ Automatic specialist selection based on request analysis
✅ Transparent communication of which expertise is being applied
✅ Seamless integration with existing workflows
✅ Override capability for explicit agent mentions
✅ Fallback to orchestrator for complex tasks
Result: User gets specialist-level responses without needing to know the system architecture.
Next Steps: Integrate this skill into GEMINI.md TIER 0 rules.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

