We use cookies to enhance your browsing experience, analyze site traffic, and personalize content. By clicking "Accept All", you consent to our use of cookies. You can customize your preferences or reject non-essential cookies.
Learn more about our cookie policySkills are markdown-based instruction sets you attach to models or invoke in chat. Teach your AI how to approach tasks — code review, writing style, troubleshooting — without writing code.
Skills are reusable, markdown-based instruction sets that teach AI models how to approach specific tasks. Unlike Tools (which run executable Python code), Skills are plain-text instructions — guidelines, playbooks, or rules that the model follows when generating responses. Think of them as "how-to" guides you attach to a conversation or a model. When the model needs to perform a task, it can load the relevant skill and apply its instructions.
Skills vs Tools
Tools execute code (Python, APIs) — they run in a sandbox and can call external services. Skills are instructions only — no code execution, no security risk. They're safe to create and share, and they keep your context window efficient.
Skills use a smart lazy-loading design so the model's context window stays efficient:
Manifest injection
When a skill is active (bound to a model or mentioned in chat), only a lightweight manifest with the skill's name and description is injected into the system prompt. The model sees what skills are available without loading full content.
On-demand loading
The model receives a view_skill built-in tool. When it decides it needs a skill's full instructions, it calls view_skill with the skill name to load the complete content. Only the skills the model actually uses get loaded into context.
Type $ in the chat input to open the skill picker. Select a skill and it attaches to your message (like @ for models or # for knowledge). The skill manifest is injected for that conversation.
Go to Workspace → Models, edit a model, and scroll to the Skills section. Check the skills you want that model to always have. When you chat with that model, the selected skills' manifests are automatically injected.
Here's a simple skill you could create to improve code review responses:
Skill: code-review-guidelines
1. Check for correctness and logic errors. 2. Look for security vulnerabilities (SQL injection, XSS). 3. Verify error handling and edge cases. 4. Suggest performance improvements. 5. Ensure code follows project style guidelines.
When you ask the model to review code, it loads this skill and applies these steps — giving you consistent, thorough reviews every time.
Code review guidelines
Step-by-step rules for thorough, consistent code reviews
Writing style rules
Tone, format, and structure for emails, docs, or reports
Troubleshooting playbooks
Systematic steps for debugging or incident response
Domain-specific instructions
Legal, medical, or financial compliance guidelines
Response formatting
How to structure answers (bullet points, sections, examples)
Per-model behavior tuning
Customize how a specific model approaches tasks
Skills work alongside other AI·Collab features. Explore these related guides:
Outgoing: Admin/user webhooks notify Slack/Discord. Incoming: Channel webhooks let external services POST into your channels.
Read moreDiscover how frontier AI models can now manage your memories, search knowledge bases, and access chat history—without you asking.
Read moreLearn how AI·Collab transforms PDFs into accurate AI answers — from OCR to embedding, hybrid search, and EU-hosted reranking. All data stays in Europe.
Read moreGet started today. Access models from OpenAI, Google, Anthropic, Grok and more.
GDPR compliant · Zero data retention · Cancel anytime