The one-line definition
A Skill is a packaged piece of knowledge or workflow the AI Employee can draw on. It might be three lines describing a formatting rule, a forty-page methodology document, or a callable workflow you trigger by typing /skill-name in your prompt. Skills sit on top of the platform and are reusable across every job the Employee handles.
What's inside a Skill
- A description. One or two sentences telling the Employee what the Skill is for. The Employee reads descriptions to decide which Skill applies to the current job.
- Instructions. The actual content — how the work should be done, what to check, what to avoid, what good output looks like.
- Reference files (optional). Playbooks, compliance frameworks, pricing matrices, sample outputs. The Agent loads these when the Skill is in use.
- Inputs (optional). Parameters you pass when invoking the Skill —
/close-package North Americaand/close-package Europerun the same workflow with different scope.
How Skills get invoked
Two ways. First, the Employee picks them automatically: it reads each Skill's description and pulls the relevant ones into context when the work matches. Second, you invoke them explicitly — type / to open the slash menu and pick the Skill you want. Explicit invocation is useful when you know exactly which workflow should run; automatic invocation is useful for everything else.
What Skills are good for
- Encoding institutional knowledge. The way your team actually does things — your standards, judgment calls, process docs — packaged so every AI Employee on the team works from the same playbook.
- Repeatable structured workflows. Quarterly review, pricing approval, support-ticket triage — anything that has a shape you'd otherwise re-explain at the top of every run.
- Domain expertise. Legal review checklists, financial modeling frameworks, healthcare interoperability patterns. The Employee inherits the expertise instead of guessing.
- Sharing across the team. Once one person writes a Skill that works, the whole team installs it and dispatches Agents that all run the same way.
Skills vs other CLI primitives
Skills are knowledge and workflow. Sub-agents are execution capacity. MCP tools are external integrations. Hooks are deterministic policy. Slash commands are typing shortcuts. The right design usually combines two or three — a Skill that calls MCP tools and dispatches sub-agents, with hooks enforcing approval before any destructive action.
Where to write your first Skill
Inside Syntic Forge or Syntic CLI, create a new Skill, write the description, paste in the instructions, drop in any reference files, and save. The Employee picks it up immediately and uses it on the next job that matches. Most teams start with one — the workflow they're tired of re-explaining — and grow from there.