Custom subagents

A subagent is a specialized version of Amara that you define once and reuse across sessions. Each subagent carries its own system prompt, its own set of allowed tools, and optionally its own model choice. When Syntic Code delegates a task to a subagent, that subagent works in a fresh context window and returns only its final answer — keeping your main conversation focused and uncluttered.

Subagents shine when you have a recurring kind of work: a reviewer that always checks for security issues, a test-writer that follows your house style, or a documentation author that knows your product’s voice. Instead of re-explaining the role every time, you encode it once and call on it whenever you need it.

Defining a subagent

Subagents live as Markdown files in .syntic/agents/ inside your project (for team-shared agents) or in ~/.syntic/agents/ (for personal agents available everywhere). Each file uses YAML frontmatter to describe the agent, followed by the system prompt in the body:

---
name: security-reviewer
description: Audits diffs for injection, auth, and secrets issues
tools: Read, Grep, Bash
model: syntic-pro
---
 
You are a focused security reviewer. Examine only the changed code.
Flag concrete, exploitable issues with file and line references.
Do not comment on style. Return a ranked list of findings.

The name is how you invoke the agent, description helps Amara decide when to delegate automatically, tools is an allowlist (omit it to inherit the full toolset), and model pins a specific Syntic model.

Invoking and iterating

You can call a subagent explicitly — “use the security-reviewer on this branch” — or let Syntic Code pick one automatically when a task matches its description. Keep prompts sharp and scope tools tightly: a reviewer rarely needs write access, and a narrow toolset makes an agent both faster and safer. Store project subagents in version control so your whole team inherits the same specialists, and refine the system prompts as you learn what produces the best results.