CLI tool for scaffolding AI agent specifications
npm install @aptove/agentspecCLI tool for scaffolding AI agent specifications. Create structured specs that AI tools can use to generate deployment artifacts.
``bash
npm install -g @aptove/agentspec
Quick Start
`bash
Create a new agent spec project
agentspec init my-agentNavigate to the project
cd my-agentEdit the spec files in specs/
Then validate your specs
agentspec validateWhen ready, publish to agent-catalog
agentspec publish
`Commands
$3
Initialize a new agent spec project with interactive setup.
`bash
agentspec init # Interactive mode
agentspec init my-agent # With name
agentspec init -t local-agent # With template
agentspec init --no-interactive # Use defaults
`Options:
-
-t, --template - Template to use (minimal, local-agent, cloud-agent)
- --no-interactive - Skip prompts, use defaults$3
Validate agent spec files for completeness and correctness.
`bash
agentspec validate # Validate ./specs
agentspec validate ./path # Validate specific path
agentspec validate --strict # Strict mode (all warnings are errors)
`Options:
-
--strict - Enable strict validation mode$3
Publish agent spec to the agent-catalog repository.
`bash
agentspec publish # Interactive publish
agentspec publish --dry-run # Show what would happen
agentspec publish -y # Skip confirmation
`Options:
-
--dry-run - Show what would be published without making changes
- -y, --yes - Skip confirmation promptsSpec Format
The generated spec structure:
`
specs/
├── agent.md # Main agent definition (name, type, protocols)
├── environment.md # Runtime environment (OS, containers, resources)
├── components.md # Tools, MCP servers, credentials
├── capabilities.md # What the agent can do
└── generation.md # Instructions for AI to generate artifacts
`$3
Specs use special markers to separate human notes from AI-readable content:
`markdown
Notes for humans only - AI will skip this
Content for AI to read and process
`Workflow
1. Scaffold - Run
agentspec init to create spec structure
2. Customize - Edit spec files manually and with AI assistance
3. Validate - Run agentspec validate to check completeness
4. Generate - Use external AI (Copilot, Gemini, etc.) to generate artifacts
5. Publish - Run agentspec publish` to add to agent-catalogMIT