Mistral Vibe: Slash-Command Skills for Custom Developer Workflows
Mistral VibeMistral Vibe 2.0 introduces user-defined slash commands through a skills system, enabling developers to create reusable prompt templates for common tasks like deployments, linting, and documentation generation. Skills are defined as Markdown files with YAML frontmatter and are automatically discovered from project directories, appearing in the autocompletion menu alongside built-in commands.
Sources & Mentions
2 external resources covering this update
Slash-Command Skills Bring Extensibility to Mistral Vibe
Alongside subagents, Mistral Vibe 2.0 introduced a skills system that allows developers to define custom slash commands for frequently used workflows. This feature transforms Mistral Vibe from a general-purpose coding agent into a customizable platform that adapts to each team's specific development patterns.
How Skills Work
Skills are user-defined prompt templates stored as Markdown files with YAML frontmatter in a SKILL.md file. When Mistral Vibe discovers these files in configured directories, it registers them as slash commands that appear in the autocompletion menu alongside built-in commands.
For example, a team could define a /code-review skill that instructs the agent to check for security vulnerabilities, style violations, and test coverage, or a /deploy skill that walks through a standard deployment checklist.
Skill Discovery and Configuration
Skills are automatically discovered from .vibe/skills/ directories (and as of v2.2.0, also from .agents/skills/ following the Agent Skills standard). Configuration is managed through config.toml with three modes:
- Whitelist mode: Only explicitly listed skills are available when
enabled_skillsis set. - Blacklist mode: All discovered skills are available except those in
disabled_skills. - Mixed mode: When both are set,
enabled_skillstakes precedence.
Skills support filtering via glob patterns, regex, and exact matching, giving teams fine-grained control over which workflows are available in different project contexts.
Integration with the Agent System
When a slash command is invoked, the skill's content is injected into the conversation as a user message, providing the LLM with the structured instructions defined in the skill template. This approach keeps the system prompt clean while allowing rich, context-specific instructions to be loaded on demand.
Practical Applications
Common use cases for skills include code review checklists, documentation generation templates, test writing workflows, debugging procedures, and performance refactoring guides. Teams can share skill definitions through version control, creating a standardized library of AI-assisted workflows.