.----------------. | _ _ | | | | | | | | | | __ | | | "Code responsibly. | | |__/ \__| | | Or don't." | |____________| | '----------------'
The Claude sUITE Framework
Orchestrate teams of Claude Code agents. Turn complex development tasks into seamless, multi-agent sessions where the only thing getting hammered is your technical debt.
Features
Auto context-switches between specialized personas — Plan, Build, Improve, and Question — based on file paths and domain detection.
Six enforcement hooks run on every agent. Supply chain verification, network egress whitelisting, prompt injection detection, and scope enforcement.
Agents don't just work — they learn. Every build cycle feeds the review-improve loop, updating tips.md and expertise.yaml automatically.
Five commands cover every workflow. /do for single-domain tasks, /do-teams for parallel work, /improve for maintenance.
Quick Start
# Run from your project root
curl -fsSL https://cuite.quest/install.sh | bashOr manually:
cd /path/to/your-project
git clone https://github.com/fentas/cuite.git /tmp/cuite
/tmp/cuite/bin/cuite init
rm -rf /tmp/cuiteThen bootstrap your domains:
/cuite-initThis scans your repo, detects languages and sub-projects, proposes domains, and on approval generates domains.md, domain-map.conf, and starter expertise files.
Architecture
Framework travels via git subtree. Symlinks bridge to where Claude Code expects files. Your domain expertise stays in your project.
.claude/cuite/.claude/.claude/