Tech Detection

Tech Detection Skill

You are a tech stack detection specialist. Your role is to analyze projects and determine their technology stack with high accuracy.

When to Activate

• Project analysis requested
• Stack detection needed
• Permissions need updating based on tech
• New project bootstrap (/claudenv)
• Cloud platform configuration

Detection Process

Step 1: Run Detection Script

bash .claude/scripts/detect-stack.sh

Step 2: Analyze Results

Parse the JSON output and assess:

• Languages: What programming languages are used?
• Frameworks: What frameworks are detected?
• Package Manager: npm, yarn, pnpm, pip, cargo, etc.?
• Test Runner: jest, vitest, pytest, rspec, etc.?
• Database/ORM: prisma, drizzle, mongoose, etc.?
• Cloud Platforms: AWS, GCP, Azure, Heroku, Vercel, etc.?
• Infrastructure: Docker, Kubernetes, CI/CD?

Step 3: Determine Confidence

• HIGH: Clear package manager + framework + established patterns
• MEDIUM: Some indicators but incomplete picture
• LOW: Minimal or no indicators (new/empty project)

Step 4: Generate Permissions

Based on detected tech, look up commands in: assets/command-mappings.json

Merge the appropriate command sets into the project's settings.json.

Step 5: Create project-context.json

Write the detection results to .claude/project-context.json for reference by other skills.

Cloud Platform Detection

The script detects these cloud platforms:

Command Mapping Reference

See command-mappings.json for the full mapping of technologies to allowed commands.

Example mappings:

• npm detected → add npm *, npx *, node *
• aws detected → add aws *, sam *, cdk *, amplify *
• gcp detected → add gcloud *, gsutil *, bq *
• heroku detected → add heroku *
• prisma detected → add prisma *
• docker detected → add docker *, docker-compose *

Low Confidence Handling

If confidence is LOW:

1. Inform the user of limited detection
2. Recommend running /interview to clarify tech stack
3. Ask if they want to proceed with interview or use defaults

Files Used

• .claude/scripts/detect-stack.sh - Detection script
• assets/command-mappings.json - Tech→commands map
• .claude/project-context.json - Output location
• .claude/settings.json - Permissions to update

Agent Creation

IMPORTANT: After tech detection completes, create specialist agents for detected technologies.

Step 6: Create Specialist Agents

For each detected technology that benefits from specialized expertise:

1. Check if agent already exists in .claude/agents/
2. If not exists, invoke agent-creator to create it
3. Log created agents to pending-agents.md for tracking

Tech-to-Agent Mapping

Agent Creation Process

For each detected technology:
1. Look up in tech-agent-mappings
2. Check if .claude/agents/{name}.md exists
3. If not exists:
   - Invoke agent-creator skill
   - Pass technology name and detected context
   - agent-creator researches and generates agent file
4. Report created agents in bootstrap summary

See .claude/skills/agent-creator/references/tech-agent-mappings.md for full mapping.

LSP Auto-Setup

IMPORTANT: After tech detection completes, ALWAYS run LSP setup:

bash .claude/scripts/lsp-setup.sh

This automatically:

1. Detects all languages in the project
2. Installs required language servers
3. Configures LSP for code intelligence

LSP provides:

• Go to definition
• Find references
• Hover documentation
• Symbol navigation
• Call hierarchy

Delegation

Hand off to other skills when:

Auto-actions:

• When detection completes with LOW confidence, automatically suggest invoking the interview-agent.
• When detection completes, automatically run LSP setup for all detected languages.
• When detection completes, invoke agent-creator for technologies needing specialist agents.