langfuse
Monitor LLM Apps with Langfuse
LLM teams need visibility into traces, cost, latency, quality, and prompt changes. This skill guides Claude, Codex, and Claude Code users through practical Langfuse instrumentation and evaluation workflows.
Install with my Agent
Copy this request to your Agent. It includes the canonical Skill page and manifest.
Review the Skillstore skill "langfuse" from https://skillstore.io/skills/sickn33-langfuse.md and its manifest at https://skillstore.io/api/skills/sickn33-langfuse/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.Your Agent should still show its plan and request any confirmation required by the security policy.
Agent-readable resources
Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.
Test it
Using "langfuse". I need to add observability to a support chatbot.
Expected outcome:
The skill would recommend creating a trace per user request, logging each model generation, attaching user and session identifiers, recording token usage, adding support-specific metadata, and flushing before serverless exits.
Using "langfuse". Our LangChain agent is hard to debug in production.
Expected outcome:
The skill would suggest a Langfuse callback handler, meaningful trace and span names, agent-level metadata, callback configuration for chains and tools, and focused tags for production investigations.
Using "langfuse". We want to compare prompt versions safely.
Expected outcome:
The skill would propose prompt version tracking, feedback and evaluation scores, datasets for regression checks, cost and latency monitoring, and tags that separate experiments from production traffic.
Security Audit
SafeThe static findings are false positives from markdown code fences, placeholder credential examples, and normal Langfuse service URLs. I found no prompt injection, data exfiltration intent, unauthorized command execution, or system reconnaissance in SKILL.md.
Risk Factors
⚙️ External commands (7)
🌐 Network access (2)
🔑 Env variables (2)
Share & cite this report
Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.
Copy report link
https://skillstore.io/skills/sickn33-langfuse/audits/4?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdown badge
[](https://skillstore.io/skills/sickn33-langfuse?utm_source=security_passport_badge)HTML badge
<a href="https://skillstore.io/skills/sickn33-langfuse?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/sickn33-langfuse/security.svg" alt="Skillstore security assessment" loading="lazy"></a>Embed card
<iframe src="https://skillstore.io/embed/skills/sickn33-langfuse.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>Academic citations (APA · BibTeX · CFF)
APA citation
sickn33. (2026). langfuse security audit report (audit version 4) [Author version unspecified]. Skillstore. https://skillstore.io/skills/sickn33-langfuse/audits/4BibTeX citation
@techreport{sickn33-sickn33-langfuse-2026,
author = {sickn33},
title = {langfuse security audit report (audit version 4)},
institution = {Skillstore},
year = {2026},
number = {4},
url = {https://skillstore.io/skills/sickn33-langfuse/audits/4},
note = {Author version unspecified}
}CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "langfuse security audit report (audit version 4)"
version: "unspecified"
type: report
authors:
- name: "sickn33"
date-released: "2026-07-07"
url: "https://skillstore.io/skills/sickn33-langfuse/audits/4"
identifiers:
- type: other
value: "skillstore:sickn33-langfuse:audit:4"
description: "Skillstore immutable audit report identifier"
Skillstore Score
Why this score Evidence Confidence: HighWhat You Can Build
Instrument a Chat Application
Add traces, generations, user identifiers, session identifiers, and token usage to an LLM chat workflow.
Monitor Production Quality
Use scores, tags, metadata, and sessions to investigate latency, cost, user feedback, and prompt regressions.
Trace Agent Frameworks
Connect Langfuse callbacks to LangChain chains, agents, and retrievers for end-to-end observability.
Try These Prompts
Use the Langfuse skill to outline the minimum tracing setup for my chat application. Include traces, generations, user identifiers, session identifiers, usage, and flush behavior.
Review my planned Langfuse instrumentation for missing trace fields, weak metadata, noisy spans, and serverless flush risks. Give practical corrections.
Design a Langfuse scoring approach for customer support conversations. Include feedback scores, automated quality signals, tags, metadata, and dataset usage.
Create a Langfuse observability architecture for a LangChain agent system with OpenAI calls, retrievers, streaming, async execution, prompt versions, and regression tracking.
Best Practices
- Use stable user and session identifiers so traces can be grouped and debugged.
- Log only useful spans, metadata, tags, and scores to keep observability data searchable.
- Flush traces at the end of serverless or short-lived jobs to prevent data loss.
Avoid
- Do not trace every internal step when it creates noise without diagnostic value.
- Do not omit user and session context from production traces.
- Do not rely on dashboards alone without scores, datasets, or feedback loops.