Skills

Skills are composable, reusable agent behaviors. A skill bundles instructions, optional scripts, and reference files that get loaded into the agent’s context when relevant.

Think of a skill as a domain expert you can install — “How to write a migration,” “How to ship a Stripe webhook,” “How to debug a flaky CI test.”

Anatomy of a Skill

A skill is a directory containing a SKILL.md file plus optional scripts and references:

~/.syntic/skills/
  writing-migrations/
    SKILL.md           ← required
    scripts/
      check-locks.sh
    references/
      postgres-row-locks.md

`SKILL.md` Frontmatter

---
name: writing-migrations
description: Use when authoring database migrations to ensure online safety, backfill correctness, and rollback strategy
---
 
# Writing Migrations
 
Always write migrations that are safe for online execution...

The description is what the agent reads at session start. Make it specific — it’s how the agent decides whether to load the skill.

Installing Skills

syntic skills install writing-migrations
syntic skills install @acme-corp/internal-deploy-runbook
syntic skills install ./my-local-skill/

Skills resolve from three sources, in order:

Local directory (./path/to/skill)
Syntic Marketplace (name or @org/name)
Git URL (https://github.com/...)

Authoring a Skill

syntic skills create my-skill

This scaffolds a skill folder with a starter SKILL.md. Edit it, test by running syntic in a project where the skill is relevant, then publish:

syntic skills publish

Best Practices

One concern per skill. “Writing migrations” not “Database stuff.”
The description is the gate. If the agent doesn’t load it when needed, the description is too vague.
Lead with the rule, then the why. Future agents need to judge edge cases.
Include verification. A skill that says “do X” should also say “verify with Y.”
Reference files are searchable. Put long docs under references/, not inline.

Marketplace

Browse and publish at syntic.ai/marketplace. Authors keep 70% of paid skill revenue.

Sub-Agents — Skills run inline; sub-agents run in isolation
Hooks — Skills are content; hooks are behavior
CLI Reference — syntic skills — Command syntax

Cloud Desktop Hooks