Install
npx skills add langfuse/skills --skill "langfuse"Who it's for
This skill is for people already using Langfuse who want a coding assistant that can look up traces, manage prompts, and inspect datasets and scores. The official description teaches the assistant how to use this open-source LLM engineering platform, covering tracing, prompt management, and evaluation.
Platform overview:
https://langfuse.com/
It is not a complete tutorial for the Langfuse dashboard. It is better suited to people who already have an account and can provide an API key, and who want the assistant to query or manage objects through the official API. If you do not have an account yet, create one first, then install the skill.
Main skill documentation:
https://github.com/langfuse/skills/blob/main/skills/langfuse
Make your first attempt in 3 minutes
First confirm that you have a Langfuse account, either cloud-based or self-hosted, and that you have prepared an API key. Do not paste the key into the conversation or use a placeholder key.
Cloud:
https://cloud.langfuse.com/
Self-hosting documentation:
https://langfuse.com/docs/deployment/self-host
There is only one official installation command:
npx skills add langfuse/skills --skill "langfuse"
This installs the main skill, langfuse. Do not add a second installation command.
After installation, do a low-risk read: ask the assistant to consult the documentation for this skill or query existing traces; do not delete datasets or change production prompts yet. The page says that after installation, the assistant will automatically use these skills for relevant tasks.
A success signal is that the assistant starts discussing traces, prompts, datasets, or scores through the Langfuse API and checks the documentation instead of inventing an interface that was never mentioned.
Use cases
Case one: consult the documentation and verify existing traces. The evidence says that the main skill can query traces through the Langfuse API and find documentation. For the first attempt, read one existing trace only, or ask the assistant to open the official documentation to verify fields. Confirm the project before sending an API request.
Case two: view prompts, datasets, or scores. The evidence says that the same skill can also manage prompts, datasets, and scores. The first time, query a list or a single record, and leave “manage/modify” until after you confirm. The excerpt does not specify exact CLI subcommands; do not invent missing parameters.
Usage boundaries
Without an account and API key, you cannot use the interface. Do not put the key in a command or repository. For the first attempt, use the npx command above.
The assistant will automatically use the skill for relevant tasks, but it does not replace your permission judgment. The excerpt does not give a variable name for the key, so do not add a placeholder. Both cloud and self-hosted setups are valid account sources; do not install a second set of skills for this.