Context-optimized agent dispatch for Claude Code - run parallel agents in isolated tmux sessions
- Shell 57.9%
- Python 42.1%
- Remove Task from allowed-tools to enforce tmux isolation - Replace dev-verify.sh with tldr-powered verification - tldr diagnostics for type checking + linting - tldr dead for dead code detection (warnings) - tldr change-impact --run for selective testing - Fallback to traditional tools when tldr unavailable - Add tldr exploration section to agent prompts - tldr structure for codebase layout - tldr search for code discovery - tldr impact for pre-refactoring analysis - Update super.md documentation with new workflow Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> |
||
|---|---|---|
| .claude | ||
| docs | ||
| lib | ||
| scripts | ||
| install.sh | ||
| LICENSE | ||
| README.md | ||
| uninstall.sh | ||
claude-super-dispatch
Context-optimized agent dispatch for Claude Code with built-in dev feedback loops.
Key Feature: Dev Feedback Loop
Every agent includes a verify-fix loop (default 7 iterations):
- IMPLEMENT the task
- VERIFY with dev-verify.sh
- If FAIL: FIX and repeat
- If PASS: Signal DONE
Auto-detects: Python, Node.js, Go, Rust
Quick Start
git clone https://github.com/pookNast/claude-super-dispatch.git
cd claude-super-dispatch
./install.sh
source ~/.bashrc
Usage
/super be Add rate limiting to API
/super py Refactor pipeline --max-iter 10
/super Fix the login bug
Agent Types
| Alias | Agent | Use For |
|---|---|---|
| be | backend-developer | APIs, databases |
| fe | frontend-developer | UI, React |
| py | python-pro | Python scripts |
| db | debugger | Bug investigation |
| dev | devops-engineer | Deploy, docker |
| sec | security-engineer | Security audit |
| test | test-automator | Tests |
| api | api-designer | API design |
Status
/tmux-orchestrator:status
tmux capture-pane -t agent-{id} -p
Requirements
- Claude Code CLI with MCP tmux tools
- Python 3.8+
- tmux 3.0+
License
MIT