Custom Slash Commands for Antigravity 2.0

Github Repo: https://github.com/10xroadmap/custom-... *Introduction* Slash commands are text-driven shortcuts used to manage AI coding assistants and configure settings. They streamline the workflow by removing the need for complex, multi-step inputs. *Advantages* Commands increase efficiency by offering a quick way to operate AI coding tools in the terminal. They provide granular programmatic control over the AI agent’s operational environment. *Supported Slash Commands* */model:* Selects the default reasoning model for persistent sessions. */skills:* Allows browsing local and global encapsulated agent workflows. */mcp:* Opens a panel to configure Model Context Protocol servers. */perms:* Selects the agent autonomy level (request-review, always-proceed, or strict). *Custom Slash Commands* Custom commands help build smart shortcuts by encapsulating multi-line instructions. They standardize engineering practices through consistent, shareable command sets. They allow for dynamic context setting, such as changing a persona or output schema. *Every Slash Command is a Skill* Slash commands and skills share a unified architecture and a modular context framework. Both utilize progressive disclosure to ensure the AI loads only pertinent instructions when required. Skills can be explicitly triggered by users through direct manual input, like commands. *So What is the difference?* *Skills:* They are contextual and automatically invoked by the AI agent for complex, multi-step workflows or domain expertise. *Slash Commands:* They are manual and direct, invoked by users for quick, atomic actions like setting schedules or approving actions. *But What are Skills?* Agent Skills represent an open-standard framework that broadens an agent’s functional range. A skill is defined as a directory containing a `SKILL.md` file that provides instructions, best practices, and supplemental assets. *Creating a Slash Command* *Step 1:* Create a folder for your skill in the `.agents/skills` directory of your workspace. *Step 2:* Define the skill capabilities by creating a `SKILL.md` file in that new directory. *Structure of SKILL.md file* The file must start with Metadata (name and description) defined between triple-dashes for agent discovery and routing. The content section includes instructions for the agent to analyze parameters, apply logic, and return formatted output. *Progressive Disclosure* *Discovery:* The agent first sees a list of available skill names and descriptions when a conversation starts. *Activation:* If a skill is relevant to the task, the agent reads the full `SKILL.md` content. *Execution:* The agent follows the skill's instructions while working on your task. *Conclusion* Slash Commands are best for manual oversight and are human-triggered shortcuts for specific tasks. Skills are best for AI delegation, defining instructions and resources the agent autonomously weaves together for complex procedural steps.