What problem does it solve? When multiple AI agents work on the same project independently, they guess conventions separately and produce inconsistent output: different naming styles, duplicated utilities, and conflicting patterns. This Skill eliminates that divergence by loading a shared layer of project conventions, coding standards, and architecture decisions into every agent session. ## Core Features & Use Cases - Static Shared Context: Maintains an orchestrator-managed .opencode/shared-context.json file containing project structure, naming conventions, existing patterns, and architecture decisions that every agent reads before starting work. - Learned Conventions Memory: Agents append project-specific discoveries to an append-only .opencode/conventions.jsonl file with dedup, tagging, and supersede rules, so knowledge persists across sessions. - Multi-Agent Coordination: Defines read/write responsibilities for coder, designer, researcher, reviewer, explorer, and orchestrator agents, plus conflict resolution rules and automatic consolidation when entries exceed limits. - Use Case: A team runs parallel coder and designer agents on a Next.js project. Each session loads the shared context, so both agents use kebab-case files, Tailwind styling, and existing utilities instead of inventing divergent patterns. ## Quick Start Load the shared project context and learned conventions from the .opencode directory before starting this agent session.
