TRAE now officially supports Agent Skills.
With Skills, your agents can understand professional domain knowledge, follow organizational standards, and automate repeated workflows on demand.
We've put together a guide on how to create, import, and use Skills in TRAE.
When to Use Skills
TRAE Skills are built on the open agent skills standard. Skills are the best when you need reusable, on-demand capabilities that your agent applies automatically. Before creating a Skill, ask yourself: "Will I need this more than once, and does it work better structured instructions?"
Use Skills when you:
- Find yourself typing the same prompt repeatedly
- Need consistent outputs across multiple sessions
- Have multi-step workflows you run regularly
- Want to encode domain expertise (coding standards, brand guidelines, testing conventions, professional analysis)
Skills vs. Other TRAE Features
TRAE offers multiple ways to guide your agent. Here's how to choose:
Prompt vs Skill
Prompts are one-off conversations. If you've typed the same prompt three times, it's time to make it a SKILL.md
Rules vs Skill
Rules are always-on preferences that load globally and consume context continuously.
If your Rules file is getting bloated, consider moving workflow specific instructions into Skills.
Keep Rules for lightweight preferences like coding style or language.
Context vs Skill
Context are shared knowledge in your workspace. It is always loaded once the conversation starts and consumes context window continuously.
Skills are structured instructions that are only invoked on demand. It's best for providing instructions for repeated workflows and action-oriented behaviors.
Subagent vs Skill
Subagents are specialized workers, but Skills are portable and reusable across different agent and tools.
If multiple subagents need the same expertise, don't duplicate instructions—create a Skill that any agent can invoke.
The key difference:
Rules, Context, and Subagent instructions all consume your context window continuously. Skills only load when invoked, therefore saving tokens and keeping your agent focused.
How to build and use Skills in TRAE
First, let's take a look at the structure of a Skill.
A Skill is simply a folder containing a SKILL.md file: a markdown file that tells your agent what the skill does and how to use it. Think of it as a user manual your agent reads on demand.
Skills are written in standard Markdown format, making them easy to create, read, and share.
your-skill/
├── SKILL.md # Required: Core instructions for your agent
├── examples/ # Optional: Example inputs/outputs
│ ├── input.md
│ └── output.md
├── templates/ # Optional: Reusable templates
│ └── component.tsx
└── resources/ # Optional: Reference files, scripts, or assets
└── style-guide.mdImport community built Skills
TRAE Skills are built on open agent skills standards, fully compatible with the community ecosystem. You can find existing Skills on GitHub and import them directly.
How to import:
- Find a Skill from the community (e.g., Example Agent Skills)
- Download the folder containing
SKILL.md - Import the
SKILL.mdinto TRAE from Settings -> Rule & Skills -> Skills -> Create
You can use them as-is, or customize to fit your needs.
Create your own Skill
A Skill is a small bundle consisting of:
- A name that identifies the skill
- A description that explains what it does and when to use it
- An optional body of instructions for more complex workflows
That's it. For simple skills, a name and description are often enough. Add detailed instructions iteratively when your workflow requires specific steps or output formats.
Create a Skill in chat
The easiest way to create a skill is simply chatting with TRAE. Just describe what you need, and TRAE generates the Skill for you.
For example:
- "Create a skill that reviews my code for performance issues"
- "Make a skill for generating CSS components that stick with our brand guidelines"
- "Build a skill that formats my data analysis reports"
Create a Skill manually
If you prefer more control, you can create Skills manually in a well written Markdown format.
Basic SKILL.md template:
---
name: Skill Name
description: Brief description of what this skills does and when to use it
---
# Skill Name
## Description
A one-line summary of what this skill does.
## When to Use
The conditions that should trigger this skill.
## Instructions
Clear, step-by-step guidance for TRAE agent to follow.
## Examples (Optional)
Input/output examples that show expected behavior.
``` Tips for creating a Skill:
Start Simple, Then Iterate. Your first Skill doesn't need to be perfect.
- Start with a basic version
- Test it with real work
- Notice where outputs miss the mark
- Add more instructions to address gaps
- Repeat
Using Skills in TRAE
Once you've created or imported Skills, there are two ways to use them:
Directly tell TRAE which Skill to use.
This gives you precise control over when a Skill is applied.
Examples:
- "Use the codemap skill to summarize what has changed on this branch "
- "Apply my Frontend Design skill to build this component"
- "Run the CSV skill on this dataset"
Use explicit invocation when you know exactly which Skill you need and want predictable behavior.
Implicit invocate the relevant Skills.
Let TRAE decide. Based on your task and the Skill's "When to Use" description, TRAE automatically invokes the right Skill when it's relevant.
For example, if you have a Code Review Skill with the trigger "when the user asks for feedback on code," TRAE will automatically apply it when you say:
- "What do you think of this function?"
- "Can you review my PR?"
- "Any issues with this code?"
Tips for using Skills
- Explicitly trigger a Skill when you want to leave no room for interpretation. This is especially useful for complex tasks or when you need consistent, repeatable results.
- Write clear "When to Use" conditions in your SKILL.md. The better your trigger descriptions, the smarter implicit invocation becomes.
What You Can Do with Skills
For Consistent, High-Quality Outputs
When you need your agent to generate outputs that always match your exact standards.
This works great for aligning design guidelines, enforcing team standards, maintaining brand consistency, and ensuring code follows your conventions. Instead of hoping your agent remembers your preferences, build packed instructions into a Skill and get reliable results.
Example:Using frontend-design skill to redesign a website
For Automating Repeated Workflows
When you're tired of doing the same multi-step tasks manually.
You can turn test frameworks, coding conventions, common data analysis routines, or daily workflows you can't avoid into repeatable Skills. If you're doing something every once in a while, it's probably worth automating.
For Building & Sharing Skills
Skills aren't just for personal use—they're a way to scale expertise.
→ Turn established workflows into Skills so you never have to explain the process again
→ Turn team standards into Skills for knowledge sharing and consistent outputs across your team
→ Discover, build, and share community Skills that work across different agents
TRAE Skills is now fully launched in SOLO mode, and we are rolling out to IDE soon.
Try it now at www.trae.ai