Langfuse Skill Pack

Claude Code skill pack for Langfuse LLM observability -- 24 skills covering tracing, evaluation, prompt management, and production operations.

What is Langfuse?

Langfuse is an open-source LLM engineering platform providing:

• Tracing: End-to-end observability for LLM calls, chains, and agents (OpenTelemetry-based in SDK v4+)
• Evaluation: Score outputs with user feedback, automated evals, LLM-as-a-Judge, and dataset experiments
• Prompt Management: Version, deploy, and A/B test prompts with labels and variables
• Analytics: Monitor costs, latency, and token usage with built-in and custom dashboards
• Datasets: Create test sets, run experiments, and track quality regressions
• Self-hosted or Cloud: Deploy anywhere or use managed cloud at cloud.langfuse.com

Installation

/plugin install langfuse-pack@claude-code-plugins-plus

SDK Coverage

All skills cover both SDK versions:

| SDK | Packages | Status | |-----|----------|--------| | v3 (legacy) | langfuse | Maintenance mode | | v4+/v5 (current) | @langfuse/client, @langfuse/tracing, @langfuse/otel, @langfuse/openai | Recommended |

Skills (24)

Getting Started (S01-S04)

| Skill | What It Does | |-------|-------------| | langfuse-install-auth | Install SDK (v3 or v4+), configure API keys, verify connection | | langfuse-hello-world | First trace with startActiveObservation, observe wrapper, and v3 legacy | | langfuse-local-dev-loop | Hot reload dev workflow, immediate trace flush, optional self-hosted Docker | | langfuse-sdk-patterns | Singleton client, observe wrapper, startActiveObservation, error-safe tracing |

Core Workflows (S05-S06)

| Skill | What It Does | |-------|-------------| | langfuse-core-workflow-a | Trace LLM calls: OpenAI drop-in wrapper, RAG pipelines, streaming, Anthropic, LangChain | | langfuse-core-workflow-b | Evaluation: scores (numeric/categorical/boolean), prompt management, datasets, experiment runner, LLM-as-a-Judge |

Troubleshooting (S07-S10)

| Skill | What It Does | |-------|-------------| | langfuse-common-errors | Top 10 errors (401, missing traces, 429, import errors) with tested solutions | | langfuse-debug-bundle | Collect diagnostic bundle (env, SDK version, connectivity, redacted logs) | | langfuse-rate-limits | Batching optimization, exponential backoff, p-queue concurrency, sampling | | langfuse-security-basics | Credential validation, PII scrubbing, self-hosted hardening, secret scanning |

Production (S11-S12)

| Skill | What It Does | |-------|-------------| | langfuse-prod-checklist | Production config, shutdown handlers, pre-deploy verification script, full checklist | | langfuse-upgrade-migration | v3 to v4 migration: import changes, tracing API, env vars, prompt API, codemods |

Pro Skills (P13-P18)

| Skill | What It Does | |-------|-------------| | langfuse-ci-integration | GitHub Actions, prompt regression tests, experiment quality gates, prompt deployment | | langfuse-deploy-integration | Deploy with Vercel/Lambda/Cloud Run/Docker, self-hosted server, health checks | | langfuse-webhooks-events | Prompt change webhooks (HMAC verified), Slack alerts, CI/CD triggers | | langfuse-performance-tuning | Benchmarking, batch tuning, payload truncation, sampling, memory management | | langfuse-cost-tuning | Token/cost tracking, Metrics API queries, smart model routing, budget alerts | | langfuse-reference-architecture | Singleton + OTel, context propagation, cross-service tracing, multi-env, circuit breaker |

Flagship Skills (F19-F24)

| Skill | What It Does | |-------|-------------| | langfuse-multi-env-setup | Dev/staging/prod config, secret management (AWS/GCP/Vault), CI/CD integration, Zod validation | | langfuse-observability | Prometheus metrics, Grafana dashboards, alert rules, Langfuse Metrics API | | langfuse-incident-runbook | Triage script, severity classification, fallback mode, resolution procedures, post-mortem | | langfuse-data-handling | Data export (traces/scores/datasets), GDPR requests, anonymization, retention policies | | langfuse-enterprise-rbac | Roles/permissions, scoped API keys, SSO/SAML, self-hosted hardening, audit logging | | langfuse-migration-deep-dive | Cross-instance migration, data export/import, dual-write zero-downtime, validation |

Usage

Skills trigger automatically when you discuss Langfuse topics:

• "Help me set up Langfuse" -> langfuse-install-auth
• "Add tracing to my OpenAI calls" -> langfuse-core-workflow-a
• "Set up evaluation for my LLM app" -> langfuse-core-workflow-b
• "Debug this Langfuse error" -> langfuse-common-errors
• "Track LLM costs with Langfuse" -> langfuse-cost-tuning
• "Migrate from langfuse v3 to v4" -> langfuse-upgrade-migration

License

MIT