The Web4 Multiagent Workplace
A workspace for orchestrating AI agent teams using tmux, OOSH, and the WODA information processing pattern.
| Concept | Purpose |
|---|---|
| WODA | What→Overview→Details→Actions: processing petabytes with short context windows |
| OOSH | Object-Oriented Shell: death to flags, Tab = documentation |
| CMM | Capability Maturity Model: L1 (ad-hoc) → L5 (optimizing) |
| Peer Monitoring | Agents can't read their own TUI; peers read each other |
.claude/agents/ # Agent role definitions (SKILL.md files)
components/OOSH/ # OOSH framework variants (dev, prod, platform-specific)
session/ # Active session state, tasks, context files
session/woda/ # WODA session story and chapters
# OOSH commands (no flags, Tab discovers everything)
./oo new myscript # Create new script
./config list # Show configuration
./log level 5 # Set debug verbosity
# Multi-agent via tmux
otmux pane.split # Split pane
hiveMind team.status # Check all agents
claudeCode context.read <pane> # Measure peer's context health| Story | Description | Chapters |
|---|---|---|
| WODA Session Story | Learning tmux, OOSH, and multi-agent orchestration | 39 chapters |
| CMM4 Journey | Building a CMM4 context-aware Claude team | 15 chapters |
CLAUDE.md— Agent workflow, tmux setup, per-prompt checklistcomponents/OOSH/dev.claude/docs/— OOSH architecture docs
Before you assume, test. Before you wrap, try without the wrapper. Before you declare, verify.
Neither agent alone can self-care. Together, both can. That's interdependence by design.