Skip to content

Latest commit

 

History

History
 
 

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 

README.md

Claude Code Skills

These Markdown files define Claude Code slash-command skills that expose LayerLens SDK operations directly within the Claude Code CLI. Instead of writing Python scripts, developers can invoke LayerLens capabilities through natural language commands like /trace, /evaluate, or /investigate -- enabling AI-assisted quality assurance without leaving the terminal.

Prerequisites

  • Claude Code CLI installed and configured
  • The LayerLens Python SDK installed: pip install layerlens --index-url https://sdk.layerlens.ai/package
  • LAYERLENS_STRATIX_API_KEY set as an environment variable

To register the skills, add the samples/claude-code/skills/ directory to your Claude Code skill search path, or copy individual .md files into your project's .claude/skills/ directory.

Quick Start

After registering the skills, invoke the trace skill from Claude Code:

/trace

Claude Code will guide you through creating and uploading a trace interactively.

Skills

File Command Scenario Description
skills/trace.md /trace Developers instrumenting LLM calls Create, upload, and manage trace records. Guides the user through trace creation with prompts for model, input, and output data.
skills/evaluate.md /evaluate QA teams running evaluations Run evaluations against traces using specified judges. Supports selecting judges, setting thresholds, and reviewing results interactively.
skills/judge.md /judge Platform teams managing evaluation criteria Create, list, update, and delete judges. Provides an interactive workflow for defining judge criteria and testing them against sample traces.
skills/investigate.md /investigate On-call engineers debugging production issues Analyze traces for errors, latency anomalies, and cost outliers. Produces a structured investigation report with suggested remediation steps.
skills/benchmark.md /benchmark ML teams comparing model performance Run benchmarks across models and review comparative results. Supports custom task batteries and historical trend analysis.
skills/optimize.md /optimize Teams refining judge accuracy Optimize judge configurations by testing against labeled datasets and adjusting scoring parameters to improve precision and recall.

Expected Behavior

Each skill operates interactively within the Claude Code session. The skill prompts the user for required inputs, executes the corresponding LayerLens SDK operations, and presents formatted results directly in the terminal. All operations use the authenticated Stratix client and persist data to your LayerLens workspace.