CLI skill should document legacy-observations-v1s for self-hosted instances

[matteomanias]

Describe your idea or feedback

When using the Langfuse skill to query observations via langfuse-cli against a self-hosted Langfuse v3 instance (deployed on Railway), the v2 observations endpoint returns:

{"ok":true,"status":404,"body":{"message":"v2 APIs are currently in beta and only available on Langfuse Cloud","error":"LangfuseNotFoundError"}}

The skill's CLI reference (references/cli.md) doesn't mention this limitation. We had to discover the legacy-observations-v1s resource by trial and error:

# This fails on self-hosted:
npx langfuse-cli api observations list --trace-id <id> --type GENERATION --json

# This works:
npx langfuse-cli api legacy-observations-v1s list --trace-id <id> --type GENERATION --json

This came up when we were auditing which LLM models were used per trace — a common debugging task. The agent initially tried observations list, got a 404, and had to fall back.

What would the ideal outcome look like?

The CLI reference section in the skill should:

1. Note that observations list (v2) is Cloud-only as of Langfuse v3
2. Document legacy-observations-v1s list as the working alternative for self-hosted instances
3. Ideally: the skill should detect the 404 and automatically retry with the legacy endpoint, or default to the legacy endpoint when LANGFUSE_HOST points to a non-cloud URL