skills.rest
google
googleOFFICIAL
Explore all skills by this author →
Website𝕏@GoogleOSSGitHub

adk-debug

Inspect ADK agent sessions and trace event flows across adk web and adk run workflows.

21.1k|3.8k|Updated Apr 1, 2025
Search Tags:#adk#tool calls#agent debugging#event tracing#session inspection#llm traces#workflow runtime
One-click install
npx skills add https://github.com/google/adk-python --skill adk-debug-google
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: adk-debug
Source: https://github.com/google/adk-python/tree/main/.agents/skills/adk-debug
Command: npx skills add https://github.com/google/adk-python --skill adk-debug-google

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill solves the problem of diagnosing why an ADK agent or workflow behaves unexpectedly by helping you inspect sessions, event flow, tool calls, and LLM traces.

Core Features & Use Cases

  • What problem does it solve? Debugging agent behavior by inspecting session events and traces rather than guessing.
  • Core Features & Use Cases Use it to troubleshoot issues like missing events, tool-call failures, incorrect branching context, JSON-only outputs, and LLM call limits using both adk web endpoints and adk run output.
  • Common workflows Visualize and correlate session events with trace spans, inspect LLM request/response payloads, and run repeatable headless checks via CLI query mode.

Quick Start

Start by running adk web for visual session inspection, or run adk run in verbose mode for quick, headless debugging of a specific agent directory.

Frequently Asked Questions about adk-debug

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

FAQPage Schema
How do I debug missing events in an ADK agent workflow?▼

To debug tool-call failures in ADK agents, inspect LLM request and response payloads using session inspection. This reveals JSON schema misconfigurations or incorrect branching context causing tool execution errors.

How do I inspect LLM traces to resolve tool-call failures in ADK agents?▼

Inspect LLM traces to resolve tool-call failures by fetching raw session and trace JSON. Event-flow tracing pinpoints JSON schema misconfigurations or incorrect branching context causing tools to fail during execution.

Why does my ADK agent get stuck in excessive LLM-call loops?▼

ADK agents enter excessive LLM-call loops when workflow routing fails or tools return malformed outputs. Use session inspection and event-flow tracing to examine LLM request and response payloads, identifying the routing issue causing the loop.

How do I troubleshoot human-in-the-loop (HITL) pause issues in ADK workflows?▼

Troubleshoot HITL pause issues by inspecting session events and trace spans via adk web. Event-flow tracing identifies missing or hidden events that prevent the workflow from resuming correctly after human interaction.

Can I run headless debugging checks for ADK agents from the command line?▼

You can run headless debugging checks using adk run in verbose mode for a specific agent directory. This CLI query mode allows repeatable event-flow tracing and session inspection without the web interface.

Related Skills

Explore More Software Engineering Skills→
MattMaggMattMagg
3

ADK Core

Initialize Google ADK projects and agents with Python or YAML configurations.

Community
Intermediate
freedomw1987freedomw1987
2

sse-chunk-boundary-debug

Diagnose and fix JSON parse errors from SSE chunk boundary splits.

Community
Intermediate
rahulgupta2018rahulgupta2018

adk-debug

Inspect ADK agent sessions, event flows, and LLM interactions.

Community
Advanced
ricardorochericardoroche
2

observability-logging

Automate JSON-structured logging and OpenTelemetry tracing for AI/LLM pipelines.

Community
Advanced