Verified:GitHub

sessions

Diagnose Synapse Claude Code plugin session lifecycle issues by mapping hooks to state changes.

29|3|Updated Mar 20, 2026

Search Tags:#debugging #synapse #mcp tools #session lifecycle #claude code plugin #sub-agent orchestration #parallel experiments

Ready to run

npx skills add https://github.com/Vincentwei1021/Synapse --skill sessions-vincentwei1021

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: sessions
Source: https://github.com/Vincentwei1021/Synapse/tree/main/public/synapse-plugin/skills/sessions
Command: npx skills add https://github.com/Vincentwei1021/Synapse --skill sessions-vincentwei1021

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

The sessions skill helps you understand and troubleshoot how the Synapse Claude Code plugin manages session lifecycle, context injection, and multi-agent parallel execution so your experiments run reliably.

Core Features & Use Cases

Hook-by-hook session lifecycle understanding: Learn what each plugin hook does (SessionStart, UserPromptSubmit, Task spawning, SubagentStart/Stop, SessionEnd) and what local state it touches.
Predictable multi-agent orchestration: Use the Task sub-agent workflow to run multiple experiments in parallel while the plugin handles session bookkeeping automatically.
Correct recovery and debugging guidance: Diagnose common failure modes like inactive sessions, duplicate sessions, missing checkin context, and experiments stuck in in_progress.

Quick Start

Use the sessions skill to explain which hook should have created, updated, or cleaned up a specific Synapse session when a sub-agent behaves unexpectedly.

Frequently Asked Questions about sessions

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

FAQPage Schema

How does the Synapse Claude Code plugin manage session lifecycle?▼

The Synapse plugin manages session lifecycle by mapping plugin hooks like SessionStart, UserPromptSubmit, and SessionEnd to state changes and MCP calls within the local .synapse/ directory.

How do I debug inactive sessions or experiments stuck in in_progress?▼

To debug inactive or stuck experiments, trace hook-level behavior and local .synapse/ state layout to identify whether SessionStart or SubagentStop hooks correctly updated the session context and cleaned up bookkeeping.

Can I run parallel experiments using Synapse sub-agents?▼

Yes, you can run parallel experiments by using the Task sub-agent workflow, which dispatches multiple experiments while the Synapse plugin automatically handles session bookkeeping and orchestration rules.

Why is context injection missing during Synapse session teardown?▼

Missing context injection during teardown often occurs when the SessionEnd hook fails to properly trigger MCP calls or update the local .synapse/ state layout before closing the agent session.

What is the best way to diagnose duplicate sessions in Synapse?▼

The best way to diagnose duplicate sessions is to examine the hook-level behavior of Task spawning and SubagentStart to verify session creation rules and check for overlapping state changes in the local .synapse/ directory.