| name | machina-agent-builder |
| description | Discover and use this skill for create, scaffold, update, validate, analyze, trace, debug, install, and maintain requests involving Machina agents, agent templates, workflows, connectors, prompts, mappings, documents, skills, and MCP-backed template installation. |
Machina Agent Builder
End-to-end skill for building, validating, installing, debugging, and maintaining Machina agent templates and their complete resource graphs.
References
Procedural guides for each lifecycle stage:
| Reference | Description |
|---|
| init.md | Scaffold new template project from scratch |
| create.md | Generate individual YAML components |
| validate.md | Check YAML files against correct patterns |
| install.md | Import templates via MCP (local or Git) |
| analyze.md | Analyze template structure and dependencies |
| trace.md | Trace agent execution chain with variable propagation |
| secrets.md | Configure vault secrets for connectors |
| api.md | MCP operations for all entities (CRUD, execute, search) |
| yaml-reference.md | Complete YAML syntax and patterns |
| connectors.md | Current repository inventory and selected guidance |
Schemas
Schema and reference patterns grounded in current repository examples:
Intent Routing
When the user triggers this skill, load the appropriate reference:
| User Says | Reference |
|---|
| "init template", "scaffold template", "new template project" | init.md |
| "create agent", "create workflow", "update prompt", "maintain mapping", "scaffold YAML" | create.md |
| "validate", "check YAML", "verify template" | validate.md |
| "install", "import", "deploy template" | install.md |
| "analyze", "what's in this template", "overview" | analyze.md |
| "trace", "execution chain", "variable flow", "debug agent" | trace.md |
| "secrets", "credentials", "configure API key" | secrets.md |
| "API", "MCP operations", "search agents", "execute workflow" | api.md |
| "YAML syntax", "task types", "expression syntax", "foreach" | yaml-reference.md |
| "connectors", "list connectors", "connector docs" | connectors.md |
| "report bug", "found issue", "feedback", "suggest improvement" | Show Feedback & Contributing section below |
For YAML field specifics, read the relevant schema from schemas/.
Workflow
Name: machina-agent-builder-check-setup
Validates that the doc-structure document exists and returns its content.
| Input | Default | Description |
|---|
document_name | doc-structure | Document to check |
| Output | Description |
|---|
doc-structure | Document content |
check-status | Workflow execution status |
Feedback & Contributing
Found a bug, unexpected behavior, or have an improvement idea while using this skill?
- Report issues: Open an issue with the label
machina-agent-builder
- Contribute fixes: Fork the repo, apply your fix, and open a PR against
main
- What to include: Describe what you expected vs. what happened. For schema/reference fixes, include before/after YAML examples.
When the user encounters an error, unexpected scaffold output, or expresses frustration with a skill behavior, proactively mention this section and suggest opening an issue.
Key Constraints
- SDK composition only: Use existing MCP/SDK resources only. If an intent cannot be expressed with supported primitives, reject it explicitly; do not generate raw Client API or HTTP proxy code.
- Connector .py files: No helper functions outside command functions. Each function must be self-contained.
- Expression syntax: Always use
$.get('field') — never ${field} or $field
- Prompt files: Use
prompts: array (not prompt:) with instruction: (not messages:)
- Connector YAML: Use
filetype: (not type:) and filename: (not script:)
- Install order: connectors → documents → prompts → mappings → workflows → agents → skills
- Vertex connector policy: Every task connector mapping with
name: google-genai and a command must include location: global and provider: vertex_ai in that same mapping.
