Prompt and Skill Management
Prompt/skill management is a systematic approach to storing, versioning, and retrieving the instructions for your LLM application. Instead of hardcoding prompts and skills in your application code, you manage them centrally in Langfuse.
Prompts are text or chat templates your application compiles for model calls. Skills are task instructions in a SKILL.md file, with optional supporting files, that your agent loads when a task needs them. See prompts and skills at a glance to choose between them.
Watch this walkthrough of Langfuse Prompt Management and how to integrate it with your application.
Decouple prompt and skill updates from code deployment
In most LLM applications, iterating on instructions and deploying code are handled by different people. Product managers and domain experts iterate on prompts and skills, while engineers manage deployments.
With prompts and skills in code, a simple text change requires engineering involvement, code review, and a full deployment cycle, turning a 2-minute update into hours or days of waiting.
When prompts and skills live in Langfuse, non-technical team members update them directly in the UI while your application fetches the version labeled production. This separation of concerns means you ship prompt and skill updates without involving engineering or triggering a deployment.
No latency, no availability risk
Langfuse Prompt and Skill Management adds no latency to your application. Prompts are cached client-side by the SDK, so retrieving them is as fast as reading from memory. See the caching docs page for more details. Skills installed with the CLI are files your agent reads from disk, so loading them does not call Langfuse.
The SDKs do not cache skills. Each skill fetch through an SDK or the API is a request to Langfuse, so it adds network time and depends on Langfuse being available. Cache file content by its hash to reuse it across requests (fetch files on demand). Skills installed with the CLI keep their installed version until you run langfuse skills update.
Getting started
Start by adding your first prompt or creating your first skill in Langfuse, and connecting it to your application. You can create prompts and skills from scratch in the UI or import existing ones from your application.
Take a moment to understand the core concepts: prompt types, skill structure, versioning, and labels.
Once you have prompts or skills in Langfuse and are using them in your application, there are a few things you can do to get the most out of Langfuse Prompt and Skill Management:
- Link prompts and skills to traces to analyze performance by prompt version and see which skill versions your agent loaded
- Use version control and labels for prompts and skills to manage deployments across environments
Looking for something specific? Take a look under Features for guides on specific topics.
Last updated on