By epiphytic
Claude Code plugin for spawning sandboxed LLM instances with intelligent resource provisioning and lifecycle management
A Claude Code plugin for spawning sandboxed coding LLMs with intelligent resource provisioning and lifecycle management.
The infinite-improbability-drive enables you to delegate complex coding tasks to isolated LLM instances. When you spawn a sub-LLM:
# Basic spawn
infinite-improbability-drive spawn "fix the auth bug"
# Use AISP mode for clearer LLM communication
infinite-improbability-drive spawn --aisp "implement user authentication"
# Spawn a team (primary + reviewer)
infinite-improbability-drive spawn-team "implement feature X"
| Document | Description |
|---|---|
| Architecture | System design and component overview |
| Configuration | All configuration options explained |
| Watcher Agent | How the orchestration agent works |
| Document | Description |
|---|---|
| CLAUDE.aisp | Development instructions |
| architecture.aisp | Type system and data flow |
| configuration.aisp | Config schema specification |
| watcher.aisp | Agent behavior specification |
┌─────────────────────────────────────────────────────────────┐
│ Host LLM Session │
│ (receives spawn request, gets summary results) │
└─────────────────────┬───────────────────────────────────────┘
│ spawn command
▼
┌─────────────────────────────────────────────────────────────┐
│ LLM Watcher Agent │
│ - Evaluates task requirements (LLM-assisted) │
│ - Provisions sandbox (git worktree) │
│ - Monitors progress (files, commits, output) │
│ - Detects errors & manages recovery │
│ - Creates PR & summary on completion │
└─────────────────────┬───────────────────────────────────────┘
│ provisions & monitors
▼
┌─────────────────────────────────────────────────────────────┐
│ Sandboxed LLM Instance │
│ - Runs in isolated git worktree │
│ - Headless, streaming mode, no session history │
│ - Limited permissions (no --dangerously-skip-permissions) │
│ - Output piped to debug files │
└─────────────────────────────────────────────────────────────┘
Create .infinite-probability/improbability-drive.toml in your project:
[spawn]
mode = "aisp" # or "passthrough"
recovery_strategy = "moderate" # "moderate", "aggressive", "interactive"
idle_timeout = 120 # 2 minutes no activity
total_timeout = 1800 # 30 minutes max
default_llm = "claude-code" # or "gemini-cli"
[spawn.permissions]
allowed_tools = ["Read", "Write", "Edit", "Glob", "Grep", "Bash"]
denied_tools = ["Task"]
max_permission_escalations = 1
[spawn-team]
coordination = "sequential" # or "ping-pong"
max_iterations = 3
reviewer_llm = "gemini-cli"
See Configuration Guide for all options.
The sandbox enforces strict isolation:
$HOME is inaccessible--dangerously-skip-permissions blocked| Strategy | Behavior |
|---|---|
| Moderate | Attempt 1 recovery before failing |
| Aggressive | Keep trying until unfixable error |
| Interactive | Ask user for each recovery decision |
| Mode | Description |
|---|---|
| Sequential | Primary LLM completes, reviewer evaluates once |
| Ping-pong | Primary and reviewer alternate until approval |
See CLAUDE.md for development instructions, or CLAUDE.aisp for LLM-optimized instructions.
# Build
cargo build --release
# Test
cargo test
# Run
./target/release/infinite-improbability-drive spawn "your task"
MIT
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
npx claudepluginhub epiphytic/infinite-improbability-drive-v0Brainstorm-to-implementation multi-LLM development workflow plugin — three-phase pipeline with teammate Claude Code instances, beads-based task tracking, and per-plan-phase nurture and secure review
Pre-configured skill management for Claude Code — discover, activate, and invoke skills from skills.sh and GitHub with security gating.
Intelligent draw.io diagramming plugin with AI-powered diagram generation, multi-platform embedding (GitHub, Confluence, Azure DevOps, Notion, Teams, Harness), conditional formatting, live data binding, and MCP server integration for programmatic diagram creation and management.
Complete creative writing suite with 10 specialized agents covering the full writing process: research gathering, character development, story architecture, world-building, dialogue coaching, editing/review, outlining, content strategy, believability auditing, and prose style/voice analysis. Includes genre-specific guides, templates, and quality checklists.
TypeScript/JavaScript full-stack development with NestJS, React, and React Native
Complete AI coding workflow system. Self-correcting memory + persistent FTS5-indexed research wikis + auto-research loop + multi-LLM council on a single SQLite store. 33 skills, 8 agents, 22 commands, 37 hook scripts across 24 events. Cross-agent via SkillKit.
Open-source, local-first Claude Code plugin for token reduction, context compression, and cost optimization using hybrid RAG retrieval (BM25 + vector search), reranking, AST-aware chunking, and compact context packets.
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.