What you get
- Enable/disable skills per workspace without deleting them
- Feature assignment to scope skills to Code Review, Incident, Assessment, or any combination
- Two creation paths: AI generation or manual writing
- Detail view with file tree and rendered markdown preview
- AI-powered editing to refine skills after creation
Prerequisites
- A CloudThinker workspace
Quick start
Create a new skill
Click New Skill and choose a creation method—write instructions or generate with AI.

Creating a skill
Click New Skill to see the available creation methods.
- Create with AI
- Write skill instructions
Describe what you need in plain language and let AI generate a complete SKILL.md for you.
- Click Create with AI
- A chat session opens with the skill creation agent
- Describe your requirements—coding standards, review rules, procedures
- The agent generates a SKILL.md with proper frontmatter and structured instructions
- Review the generated skill and click Save
Managing skills
Once created, you can manage skills from the Custom Skills page.Enable or disable
Toggle a skill on or off using the switch on its card. Disabled skills are preserved but not loaded by agents.Assign to features
Use the feature checkboxes on each skill card to control where it applies:- Code Review — applied during PR analysis
- Incident — applied during incident investigation and RCA
- Assessment — applied during infrastructure assessments
View details
Click a skill card to open the detail view. This shows a split-pane layout with the file tree on the left and rendered markdown content on the right.
Edit with AI
From the detail view, click Edit with AI to open a chat session where you can refine the skill’s instructions through conversation. The AI understands the existing content and can help you restructure, expand, or simplify it.Delete
Remove a skill permanently from your workspace. This action cannot be undone.Tips
- Start focused — a skill that does one thing well is more effective than a broad skill that covers everything loosely.
- One domain per skill — separate security rules from performance guidelines. This lets you assign them independently to different features.
- Use concrete examples — agents follow examples more reliably than abstract rules. Include sample code, expected outputs, or decision tables.
- Iterate with AI — use AI editing to refine skills over time based on review results and team feedback.
- Test before deploying — enable a skill on a single feature first and review the agent’s output before assigning it broadly.
Next steps
Skill Format
Learn the SKILL.md file structure, frontmatter reference, and best practices for writing effective instructions.
Skills Overview
Understand how skills fit into the CloudThinker platform and agent workflow.
