What problem does it solve? After running trellis init, AI assistants often do not know how the generated .trellis/ directory and platform folders (.claude/, .codex/, .cursor/, etc.) fit together, so they edit the wrong files or modify upstream npm packages instead of project-local configuration. ## Core Features & Use Cases - Architecture Reference: Explains the three-layer local Trellis model (workflow, persistence, platform integration) and which generated files are safe to edit. - Customization Routing: Maps user requests like "change the workflow", "agent did not read specs", or "add a skill" to the exact local file to modify. - Use Case: A user says "the implement agent never reads my backend conventions." The AI reads the context-loading reference, checks the task's implement.jsonl, and fixes the sub-agent context hook or agent prelude in the project. ## Quick Start Ask the AI to use the trellis-meta skill to change the Trellis workflow so that task creation is skipped for one-reply questions.