# Langfuse Agent Tracer

> Automatically queries Langfuse sessions and traces to isolate agent-utilities execution runs and diagnose spawned agent orchestration errors or performance drops.

- Skill: `knuckles-team/langfuse-agent-tracer` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add knuckles-team/langfuse-agent-tracer`
- Raw SKILL.md: https://api.skillmd.com/api/skills/knuckles-team/langfuse-agent-tracer/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: Knuckles-Team (https://skillmd.com/u/knuckles-team)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/knuckles-team/langfuse-agent-tracer

---


# Langfuse Agent Tracer Workflow

**CONCEPT:DEV-001**

Automatically queries Langfuse sessions and traces to isolate agent-utilities execution runs and diagnose spawned agent orchestration errors or performance drops.

## Steps

### Step 0: list-agent-traces [skill: langfuse-mcp]
**Agent**: `scanner-agent`
**Tools**: `rep_rm_workspace, rep_rm_git`

Retrieve lists of traces or recent sessions filtered by tags representing spawned agents or the agent orchestrator using trace_list or sessions_list actions.
Expected: `trace_list_data, active_sessions`

### Step 1: User Interaction
**Agent**: `builder-agent`
**Tools**: `rep_rm_projects`

Present a structured dashboard of agent executions, highlighting traces with warning levels, high latencies, or error logs. Prompt the user to select an execution trace for deep analysis.
Expected: `selected_trace_id, diagnosis_notes`

### Step 2: fetch-trace-details [skill: langfuse-mcp]
**Agent**: `validator-agent`
**Tools**: `rep_rm_projects, gl_pipelines`

Retrieve complete telemetry span trees, inputs, outputs, and prompt details for the selected execution trace using the trace_get action.
Expected: `trace_span_details`

### Step 3: KG Persistence [depends_on: Step 0, Step 2]
**Agent**: `validator-agent`
**Tools**: `graph_write`

Persist workflow results as nodes and edges in the Knowledge Graph.
Create appropriate typed nodes with metadata and link to existing domain entities.

## Output
- Langfuse Agent Tracer results persisted in KG
- Structured report (MD/PDF)
- Audit trail with timestamps and agent attributions

## Execution

Run this workflow as a dependency-ordered DAG. Steps with no unmet `depends_on` run in parallel; dependents run after their prerequisites complete.

- **Run first (in parallel):** Step 0 — list-agent-traces; Step 1 — User Interaction; Step 2 — fetch-trace-details
- **After level 0:** Step 3 — KG Persistence

**Execution:** If graph-os is reachable, offload the whole DAG via `graph_orchestrate action=execute_workflow` (or the `kg-delegate` skill) for true parallel/swarm execution. Otherwise execute the steps natively in dependency order: run steps with no unmet `depends_on` in parallel, then their dependents.