raindrop-debug-traces

Name: raindrop-debug-traces
Availability: InStock
Author: Raindrop

Analyze Raindrop Workshop traces to explain agent behavior and pinpoint issues.

961|60|Updated May 1, 2026

Search Tags:#debugging #explanation #workshop #trace-analysis #ai-traces #span-annotations

Ready to run

npx skills add https://github.com/raindrop-ai/cli --skill raindrop-debug-traces

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: raindrop-debug-traces
Source: https://github.com/raindrop-ai/cli/tree/main/plugin/skills/debug-traces
Command: npx skills add https://github.com/raindrop-ai/cli --skill raindrop-debug-traces

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Raindrop's tracing debugging assistant helps you understand AI agent traces from Raindrop Workshop, enabling quick insight into what happened and why.

Core Features & Use Cases

Reads traces via the raindrop MCP server's read-only tools and generates concise explanations.
Can pin verdicts and evidence to runs or individual spans when span annotations are enabled (experimental).
Useful for diagnosing where latency arises, why a tool failed, or what the LLM saw at a given step.

Quick Start

Invoke /raindrop-debug-traces in your IDE chat while a run is open to begin tracing and explanation.

Frequently Asked Questions about raindrop-debug-traces

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I debug AI agent traces to explain unexpected behavior?▼

Debug AI agent traces by reading Workshop spans and tool calls to generate concise explanations of agent behavior. This helps diagnose why a tool failed or what the LLM saw at a given step.

How does trace analysis pinpoint latency issues in AI runs?▼

Trace analysis pinpoints latency by capturing spans and tool calls from a running Workshop UI. It reads traces via read-only MCP tools to show exactly where delays arise during execution.

Do I need Raindrop Workshop running locally to analyze traces?▼

Yes, you need Raindrop Workshop running locally with an open run UI to analyze traces. The Skill uses read-only MCP endpoints to capture spans without modifying your data.

Can I pin verdicts and evidence to specific spans in a trace?▼

Yes, you can pin verdicts and evidence to runs or individual spans when span annotations are enabled. This experimental feature helps highlight specific issues within the trace.

What's the best way to diagnose why an LLM tool failed during a run?▼

Diagnose LLM tool failures by invoking trace analysis in your IDE chat while the run is open in Workshop. It reads the trace history to explain what the LLM saw and why the failure occurred.

Are there limitations when using read-only endpoints for trace debugging?▼

The main limitation is that read-only endpoints prevent any data modification during trace debugging. Additionally, pinning verdicts to spans requires the experimental span annotations feature to be enabled.