Skip to main content

Skills

Introduction​

Tuanjie AI Skills are professional capability extension modules for Tuanjie AI agents. A Skill is an independent directory that packages the instruction workflows, scripts, templates, and reference documents needed to complete a specific type of specialized task into standardized capabilities that can be automatically discovered by agents and activated on demand.

It allows basic AI to quickly possess professional domain capabilities (such as code review, API detection, deployment, automation workflows) without having to redescribe complex rules each time, and without taking up too much context.

What Are Skills?​

Skills are reusable capabilities for repeatable workflows. They typically contain:

  • Instructions: Define how to perform specific tasks
  • Scripts: Executable code
  • References: Related documentation and resources
  • Metadata: Name, description, and trigger conditions

Core Value of Skills​

  • Capability Extension: Add professional skills with one click to AI, covering scenarios such as development, testing, operations, and process automation
  • Process Standardization: Encapsulate team fixed workflows into Skills to ensure unified AI execution and stable results
  • Knowledge Sharing: Skills can be shared within teams, turning personal experience into reusable tools
  • Resource Integration: Built-in scripts, templates, and documentation are directly available to AI when activated
  • Lightweight and Efficient: By default, only loads names and descriptions, loading complete content only when needed, saving resources

Scenarios Suitable for Skills​

  • You frequently repeat a certain type of task (e.g., generate weekly reports, organize meeting minutes, export spreadsheets to XLSX/PDF, perform code review checklists)
  • You want to solidify "steps for doing things/standards/constraints" into templates so the model can call them anytime in conversations
  • You want to provide unified workflows for specific teams/projects (placed in .codely-cli/skills/ within the project)

Scenarios NOT suitable for Skills:

  • Just want to run built-in commands (e.g., /help, /chat, etc.)
  • Skill content needs to actually execute external programs/access network/read-write files—Skills itself are not an execution environment; it mainly provides instructions + templates

Managing Skills via GUI​

  1. Type '/' in the dialog box and select "Manage skills" from the dropdown menu. Or directly type "/skills".
  2. In the opened dialog, we have preset some skills for you to use directly. You can also create your own skill.
  3. After creating a new skill, start a new conversation session. The newly added skill will complete its loading when the new session starts.
  4. Type '@' in the conversation window to quickly select your newly added skill.

Difference between Project and Global when creating skills:

Project represents that the skill is used in the current project.

Global represents that the skill can be used in all projects.

Managing Skills via CLI​

Skill Discovery Paths and Priorities​

Tuanjie AI automatically loads Skills from three locations, from high to low priority:

Workspace Skills

  • Path: .codely-cli/skills/ or .agents/skills/
  • Purpose: Team shared, follows the project, highest priority

User Skills

  • Path: ~/.codely-cli/skills/ or ~/.agents/skills/
  • Purpose: Personal global use, effective for all projects

Extension Skills

  • Purpose: Comes with extension plugins, available after installing extensions

Priority Rules:

Workspace Skills > User Skills > Extension Skills

Within the same level: .agents/skills/ takes precedence over .codely/skills/

How Skills Work​

  1. Discovery: When creating a new session, Tuanjie AI scans all paths and loads Skill names and descriptions
  2. Matching: Based on your needs, AI automatically determines whether to activate a certain Skill
  3. Confirmation: Interface prompts you whether to allow activating that skill
  4. Loading: After agreeing, AI loads the complete instructions, directories, and resources of that Skill
  5. Execution: AI completes the task according to the workflow defined by the Skill, effective throughout the session

Managing Within Interactive Sessions​

Use /skills series commands in the Tuanjie AI conversation input box:

  • /skills list: List all discovered skills and their status
  • /skills enable <skill_name>: Enable skill (default applies to user level)
  • /skills disable <skill_name>: Disable skill (default applies to user level)
  • /skills reload: Rescan and reload all skills

If you need to control workspace scope, add parameter:

--scope workspace

Example:

/skills disable api-checker --scope workspace

Terminal Command Line Management​

# List all skills
codely skills list

# Link local skill directory
codely skills link /path/to/your/skills

# Link to workspace
codely skills link /path/to/your/skills --scope workspace

# Install skill (Git / local directory / .skill package)
codely skills install https://github.com/xxx/skill.git
codely skills install ./my-skill
codely skills install ./package.skill

# Install to workspace
codely skills install ./my-skill --scope workspace

# Uninstall skill
codely skills uninstall my-skill --scope workspace

# Enable/disable
codely skills enable my-skill
codely skills disable my-skill --scope user

How to Trigger and Use Skills​

  1. Ensure the skill has been discovered: Can be seen with /skills list
  2. Directly describe the task in natural language, for example: "Help me check if this API is available"
  3. AI will automatically match and prompt you to activate the corresponding Skill
  4. After confirming activation, AI will execute the task according to the Skill's built-in workflow

Creating Custom Skills​

Use the built-in skill-creator skill:

Enter directly in conversation:

create a new skill called 'api-auditor'

Tuanjie AI will automatically generate:

  • Skill directory (consistent with skill name)
  • SKILL.md main file
  • scripts/, references/, assets/ standard structure

Manual Creation​

Standard Directory Structure​
my-skill/
├── SKILL.md       # Required: Skill metadata + instructions
├── scripts/       # Optional: Scripts
├── references/    # Optional: Documentation
└── assets/        # Optional: Templates, resources
SKILL.md Format​

Uses YAML metadata + Markdown instructions:

---
name: api-auditor
description: Audit and test API endpoints. Use when user asks to check, test, or audit URL/API.
---

# API Auditor Skill Instructions

You are a professional API testing expert.

## Execution Process

1. Get the URL provided by the user.
2. Call scripts/audit.js to detect the interface status.
3. Return clear detection results and recommendations.

## Must Follow

- Must use scripts in this skill's directory.
- Must output structured results: target, status, details, recommendations.

Important Rules:

  • name must be exactly consistent with the directory name
  • description should clearly state functionality + trigger scenarios so AI can accurately match

Best Practices​

  1. Description Optimization: Write clear, specific descriptions to help AI correctly select Skills
  2. Tool Permission Management: Follow the principle of least privilege, only granting Skills the tools needed to complete tasks
  3. Modular Design: Break down complex tasks into multiple reusable Skills
  4. Version Control: Include Skills in version control for easy team collaboration
  5. Refined Documentation: Add detailed usage instructions and examples for Skills

Document Version: 1.0 Last Updated: March 24, 2026