Comparison
Skill package vs CLI + MCP: which artifact should Loreto build?
When you generate with Loreto you pick an artifact: a skill package, a CLI + MCP, or both. A skill package is instructions your agent reads; a CLI + MCP is runnable code your agent (or you) can execute directly.
Quick answer
Skill package: A skill package is a SKILL.md + references + test that your agent reads and follows. CLI + MCP: A CLI + MCP is a runnable command-line tool plus an MCP server exposing it to agents.
| Skill package | CLI + MCP | |
|---|---|---|
| Form | Instructions & know-how | Executable code + tools |
| Great for | Workflows, judgment, multi-step processes | Deterministic actions & integrations |
| Runs by | Agent following steps | Direct execution / tool call |
| Ship target | .claude/skills/ | npx / MCP client |
| Pick "Both" | Agent reads the steps | and can run the tool |
Use Skill package when
- The value is the process and judgment
- You want portability and easy editing
Use CLI + MCP when
- The value is a repeatable, deterministic action
- You want to invoke it as a command
The verdict
Choose "Both" when your agent should understand the workflow and be able to execute it. Loreto builds all three from one source.
Build the skill for your side of this
Loreto turns any source into a Claude Code skill — package, CLI + MCP, or both.