By obra
Automatically saves Claude Code session state on every Stop event by running a Python autosave script with a 30-second async timeout, preventing progress loss and enabling safe workflow restarts or continuations without manual intervention.
npx claudepluginhub jmagar/lab --plugin double-shot-latteStops "Would you like me to continue?" interruptions.
Claude Code plugin that uses Claude to judge whether Claude should continue working.
Claude Code interrupts multi-step work to ask permission even when next steps are obvious:
This breaks flow.
When Claude tries to stop, a separate Claude instance evaluates the context and decides whether continuation makes sense.
Key principle: If you can type "continue" and Claude knows what to do, the plugin continues automatically.
/plugin install double-shot-latte@superpowers-marketplace
Prerequisites: Claude Code and jq command-line tool.
Multi-File Projects
"Refactor codebase to TypeScript with strict types and comprehensive tests"
API Development
"Create REST API with authentication, CRUD operations, and test coverage"
Component Libraries
"Build React components: Button, Input, Modal, Table with TypeScript and Storybook"
Development Environment
"Set up complete dev environment: Docker, CI/CD, linting, testing, deployment"
By default, the plugin uses Claude Haiku for fast, cost-effective judgments. You can configure a different model via environment variable:
Configuration (~/.claude/settings.json):
{
"env": {
"DOUBLE_SHOT_LATTE_MODEL": "sonnet"
}
}
Available models:
haiku (default) - Fast and cost-effectivesonnet - More capable reasoningopus - Highest capabilityNote: The judge model runs on every stop attempt. Using more expensive models will increase costs proportionally.
Hook not triggering:
/plugin listUnexpected behavior:
/tmp/.claude-continue-throttle-*jq is installedMIT License - see LICENSE
Autonomous work mode — pre-flight risk assessment, dangerous-ops approval, blocker decision framework, and session-scoped stop hook. Lets Claude work continuously while you're away.
Share bugs, ideas, or general feedback.
Prevents your computer from sleeping while Claude Code works hard
Easily create hooks to prevent unwanted behaviors by analyzing conversation patterns
Self-evolving Claude Code system that learns from corrections, manages context, and improves every session
Self-evolving multi-agent system for Claude Code that enables intuitive, flow-based development without rigid planning.
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques