multimodel-dev-os 0.3.0

package/TASKS.md

@@ -0,0 +1,28 @@
+# Tasks
+
+> Lightweight task tracking for AI agents and humans.
+> Agents should check this file before starting work. Update status as you go.
+
+## Current Sprint
+
+<!-- Active work items — agents pick from here -->
+
+- [ ] null
+
+## Backlog
+
+<!-- Upcoming work — not yet started -->
+
+- [ ] null
+
+## In Review
+
+<!-- Completed work awaiting review -->
+
+- [ ] null
+
+## Done
+
+<!-- Completed and merged — keep last 10 items, archive the rest -->
+
+- [x] null

package/adapters/antigravity/.gemini/settings.json

@@ -0,0 +1,13 @@
+{
+  "//": "Antigravity adapter settings for multimodel-dev-os",
+  "//note": "Source of truth is /AGENTS.md — this file provides IDE-level overrides only",
+  "agent": {
+    "preferredModel": null,
+    "contextFiles": [
+      "AGENTS.md",
+      "MEMORY.md",
+      "TASKS.md",
+      "RUNBOOK.md"
+    ]
+  }
+}

package/adapters/antigravity/AGENTS.md

@@ -0,0 +1,29 @@
+<!-- AUTO-GENERATED REFERENCE — Source of truth: /AGENTS.md -->
+<!-- Do not edit this file directly. Edit /AGENTS.md and re-sync. -->
+
+# Antigravity Agent Instructions
+
+This project uses [multimodel-dev-os](https://github.com/rizvee/multimodel-dev-os).
+
+**Full instructions:** See `/AGENTS.md` at the project root.
+
+## Antigravity-Specific Notes
+
+- Google Antigravity reads `AGENTS.md` from the repo root
+- Additional configuration via `.gemini/settings.json` for IDE-level settings
+- Antigravity supports Knowledge Items (KI) — see `MEMORY.md` for project context
+
+## Integration Points
+
+| Feature | multimodel-dev-os File | Antigravity Equivalent |
+|---------|----------------------|----------------------|
+| Agent instructions | `AGENTS.md` | Auto-detected |
+| Project memory | `MEMORY.md` | Knowledge Items |
+| Task tracking | `TASKS.md` | Task artifacts |
+| Skills | `.ai/skills/` | Skills directory |
+
+## References
+
+- [Root AGENTS.md](/AGENTS.md)
+- [Project MEMORY](/MEMORY.md)
+- [Antigravity Settings](.gemini/settings.json)

package/adapters/antigravity/setup.md

@@ -0,0 +1,36 @@
+# Antigravity Setup Guide
+
+## Overview
+
+Google Antigravity automatically reads `AGENTS.md` from the repo root. The `.gemini/settings.json` file provides optional IDE-level overrides.
+
+## Setup Steps
+
+1. **Ensure `AGENTS.md` exists** at your project root
+2. **Copy the `.gemini/` directory** to your project root:
+   ```bash
+   cp -r adapters/antigravity/.gemini/ .gemini/
+   ```
+3. **Edit `.gemini/settings.json`** if you need IDE-level overrides
+
+## Mapping to Antigravity Features
+
+| multimodel-dev-os | Antigravity Feature |
+|-------------------|-------------------|
+| `AGENTS.md` | Automatically read as agent instructions |
+| `MEMORY.md` | Use as Knowledge Item source |
+| `.ai/skills/` | Map to Antigravity skills |
+| `.ai/config.yaml` | Reference in settings.json |
+
+## Tips
+
+- Antigravity supports planning mode — align with `TASKS.md` for task tracking
+- Use `MEMORY.md` to persist context across Antigravity sessions
+- Antigravity reads the full repo tree — keep `.gitignore` clean
+
+## Troubleshooting
+
+| Issue | Solution |
+|-------|----------|
+| Settings not applied | Ensure `.gemini/` is at project root |
+| Context missing | Add files to `contextFiles` in settings.json |

package/adapters/claude/CLAUDE.md

@@ -0,0 +1,31 @@
+<!-- AUTO-GENERATED REFERENCE — Source of truth: /AGENTS.md -->
+<!-- Do not edit this file directly. Edit /AGENTS.md and re-sync. -->
+
+# CLAUDE.md
+
+This project uses [multimodel-dev-os](https://github.com/rizvee/multimodel-dev-os).
+
+**Full instructions:** See `/AGENTS.md` at the project root.
+
+## Project Rules
+
+- Read `/AGENTS.md` for build commands, coding conventions, and boundaries
+- Read `/MEMORY.md` for architecture decisions and session history
+- Check `/TASKS.md` before starting any work
+- Follow `/RUNBOOK.md` for deployment and operational procedures
+
+## Claude-Specific Notes
+
+- Claude Code reads `CLAUDE.md` from the repo root automatically
+- This file is an adapter — the source of truth is `/AGENTS.md`
+- When making decisions, update `/MEMORY.md` with the rationale
+- When completing a task, update `/TASKS.md`
+- When handing off to another agent, write a session log to `.ai/session-logs/`
+
+## Build Commands
+
+See `/AGENTS.md` → Build Commands section.
+
+## Boundaries
+
+See `/AGENTS.md` → Boundaries section.

package/adapters/claude/setup.md

@@ -0,0 +1,35 @@
+# Claude Code Setup Guide
+
+## Overview
+
+Claude Code reads `CLAUDE.md` from the repo root automatically. This adapter provides a template that references the root `AGENTS.md`.
+
+## Setup Steps
+
+1. **Copy `CLAUDE.md` to your project root:**
+   ```bash
+   cp adapters/claude/CLAUDE.md CLAUDE.md
+   ```
+2. **Ensure `AGENTS.md` exists** at your project root with build commands filled in
+3. **No further configuration needed** — Claude auto-detects `CLAUDE.md`
+
+## How It Works
+
+Claude Code loads `CLAUDE.md` at session start. The adapter file:
+- Points Claude to read `/AGENTS.md` for full instructions
+- Includes Claude-specific behavior notes (session logging, memory updates)
+- Keeps Claude aligned with the multimodel-dev-os workflow
+
+## Tips
+
+- Claude supports nested `CLAUDE.md` in subdirectories for monorepo setups
+- Use `MEMORY.md` to persist context across Claude sessions
+- Claude can read any markdown file — point it at `TASKS.md` for task context
+
+## Troubleshooting
+
+| Issue | Solution |
+|-------|----------|
+| Claude ignores CLAUDE.md | Ensure file is at project root |
+| Context too large | Switch to Caveman Mode templates |
+| Claude doesn't follow AGENTS.md | Add explicit "Read /AGENTS.md first" to CLAUDE.md |

package/adapters/codex/AGENTS.md

@@ -0,0 +1,27 @@
+<!-- AUTO-GENERATED REFERENCE — Source of truth: /AGENTS.md -->
+<!-- Do not edit this file directly. Edit /AGENTS.md and re-sync. -->
+
+# Codex Agent Instructions
+
+This project uses [multimodel-dev-os](https://github.com/rizvee/multimodel-dev-os).
+
+**Full instructions:** See `/AGENTS.md` at the project root.
+
+## Codex-Specific Notes
+
+- Codex natively reads `AGENTS.md` from the repo root
+- No additional configuration required for basic usage
+- For sandboxed execution, ensure `RUNBOOK.md` contains setup commands
+
+## Environment
+
+Codex runs in a sandboxed cloud environment. Ensure:
+- All dependencies are listed in the appropriate manifest (`package.json`, `requirements.txt`, etc.)
+- Build commands in `AGENTS.md` work without interactive prompts
+- Environment variables are documented in `RUNBOOK.md` (never include actual secrets)
+
+## References
+
+- [Codex Documentation](https://platform.openai.com/docs/codex)
+- [Root AGENTS.md](/AGENTS.md)
+- [Project RUNBOOK](/RUNBOOK.md)

package/adapters/codex/setup.md

@@ -0,0 +1,25 @@
+# Codex Setup Guide
+
+## Overview
+
+OpenAI Codex reads `AGENTS.md` from the repo root automatically. No special configuration is needed.
+
+## Setup Steps
+
+1. **Ensure `AGENTS.md` exists** at your project root with build commands filled in
+2. **Ensure `RUNBOOK.md`** contains environment setup commands (Codex runs in a sandbox)
+3. **No adapter copy needed** — Codex reads root `AGENTS.md` directly
+
+## Tips
+
+- Codex works best with explicit, non-interactive build commands
+- Include dependency install commands in `RUNBOOK.md` → Environment Setup
+- Test your build commands in a clean environment before using with Codex
+
+## Troubleshooting
+
+| Issue | Solution |
+|-------|----------|
+| Codex ignores AGENTS.md | Ensure file is in repo root, not a subdirectory |
+| Build fails in sandbox | Add all setup steps to RUNBOOK.md Environment Setup |
+| Missing dependencies | List everything in your package manifest |

package/adapters/cursor/.cursorrules

@@ -0,0 +1,30 @@
+# AUTO-GENERATED REFERENCE — Source of truth: /AGENTS.md
+# Do not edit this file directly. Edit /AGENTS.md and re-sync.
+#
+# Cursor Rules for multimodel-dev-os project
+# Full instructions: /AGENTS.md
+
+# Project Overview
+# project: null
+# stack: null
+
+# Build Commands
+# dev:   null
+# build: null
+# test:  null
+# lint:  null
+
+# Coding Conventions
+# - See /AGENTS.md for full style guide
+# - See /MEMORY.md for project-specific patterns
+
+# Boundaries — do not modify
+# - .env, .env.local
+# - node_modules/
+# - package-lock.json
+
+# Agent Behavior
+# - Check /TASKS.md before starting work
+# - Update /MEMORY.md with decisions made
+# - Follow /RUNBOOK.md for deploy procedures
+# - Log sessions to .ai/session-logs/ when handing off to another tool

package/adapters/cursor/setup.md

@@ -0,0 +1,35 @@
+# Cursor Setup Guide
+
+## Overview
+
+Cursor reads `.cursorrules` from the project root to configure agent behavior. The adapter provides a template that references the root `AGENTS.md`.
+
+## Setup Steps
+
+1. **Copy `.cursorrules` to your project root:**
+   ```bash
+   cp adapters/cursor/.cursorrules .cursorrules
+   ```
+2. **Edit `.cursorrules`** — replace `null` values with your project's actual commands and conventions
+3. **Ensure `AGENTS.md` exists** at your project root (Cursor can reference it)
+
+## How It Works
+
+Cursor loads `.cursorrules` at session start. The adapter file:
+- Contains project rules in Cursor's native comment format
+- References root `AGENTS.md` for full details
+- Includes boundary rules (files not to modify)
+
+## Tips
+
+- Keep `.cursorrules` under 100 lines — Cursor has a context budget
+- For Caveman Mode, use an even shorter `.cursorrules` with just build commands
+- Cursor also supports `@` references to files — point it at `MEMORY.md` for context
+
+## Troubleshooting
+
+| Issue | Solution |
+|-------|----------|
+| Rules not loading | Ensure `.cursorrules` is at project root (not in adapters/) |
+| Context too long | Trim `.cursorrules` or switch to Caveman Mode |
+| Cursor ignores boundaries | Reinforce no-touch rules at the top of the file |

package/adapters/gemini/GEMINI.md

@@ -0,0 +1,31 @@
+<!-- AUTO-GENERATED REFERENCE — Source of truth: /AGENTS.md -->
+<!-- Do not edit this file directly. Edit /AGENTS.md and re-sync. -->
+
+# GEMINI.md
+
+This project uses [multimodel-dev-os](https://github.com/rizvee/multimodel-dev-os).
+
+**Full instructions:** See `/AGENTS.md` at the project root.
+
+## Project Rules
+
+- Read `/AGENTS.md` for build commands, coding conventions, and boundaries
+- Read `/MEMORY.md` for architecture decisions and session history
+- Check `/TASKS.md` before starting any work
+- Follow `/RUNBOOK.md` for deployment and operational procedures
+
+## Gemini-Specific Notes
+
+- Gemini reads `AGENTS.md` from the repo root
+- This file provides Gemini-specific behavioral overrides
+- When making architecture decisions, update `/MEMORY.md`
+- When completing a task, update `/TASKS.md`
+- When handing off to another agent, write a log to `.ai/session-logs/`
+
+## Build Commands
+
+See `/AGENTS.md` → Build Commands section.
+
+## Boundaries
+
+See `/AGENTS.md` → Boundaries section.

package/adapters/gemini/setup.md

@@ -0,0 +1,34 @@
+# Gemini Setup Guide
+
+## Overview
+
+Google Gemini (via AI Studio or API) can be directed to read `AGENTS.md` or a custom `GEMINI.md` instruction file.
+
+## Setup Steps
+
+1. **Copy `GEMINI.md` to your project root:**
+   ```bash
+   cp adapters/gemini/GEMINI.md GEMINI.md
+   ```
+2. **Ensure `AGENTS.md` exists** at your project root
+3. **When starting a Gemini session**, reference `GEMINI.md` or `AGENTS.md` as system context
+
+## How It Works
+
+Unlike Cursor or Claude, Gemini does not auto-detect instruction files. You need to:
+- Include `GEMINI.md` or `AGENTS.md` content in your system prompt
+- Or reference these files when using Gemini in an IDE that supports file context
+
+## Tips
+
+- Keep `GEMINI.md` concise — Gemini has generous context but benefits from focused instructions
+- Use Caveman Mode for API-heavy workflows where tokens cost money
+- Gemini works well with structured YAML — reference `.ai/config.yaml` for project settings
+
+## Troubleshooting
+
+| Issue | Solution |
+|-------|----------|
+| Gemini doesn't follow rules | Include GEMINI.md content in system prompt |
+| Token costs too high | Switch to Caveman Mode templates |
+| Context not persisting | Use MEMORY.md and include it each session |

package/adapters/vscode/.vscode/settings.json

@@ -0,0 +1,21 @@
+{
+  "//": "VS Code adapter settings for multimodel-dev-os",
+  "//note": "Source of truth is /AGENTS.md — this file provides editor-level settings only",
+  "editor.formatOnSave": true,
+  "editor.defaultFormatter": null,
+  "editor.tabSize": 2,
+  "editor.insertSpaces": true,
+  "files.trimTrailingWhitespace": true,
+  "files.insertFinalNewline": true,
+  "files.trimFinalNewlines": true,
+  "files.exclude": {
+    "node_modules": true,
+    ".next": true,
+    "dist": true,
+    "__pycache__": true
+  },
+  "search.exclude": {
+    "node_modules": true,
+    ".ai/session-logs": true
+  }
+}

package/adapters/vscode/setup.md

@@ -0,0 +1,40 @@
+# VS Code Setup Guide
+
+## Overview
+
+VS Code uses `.vscode/settings.json` for workspace settings. This adapter provides settings aligned with multimodel-dev-os conventions.
+
+## Setup Steps
+
+1. **Copy `.vscode/` to your project root:**
+   ```bash
+   cp -r adapters/vscode/.vscode/ .vscode/
+   ```
+2. **Edit `.vscode/settings.json`** — set your preferred formatter and tab size
+3. **Install recommended extensions** (optional):
+   - EditorConfig for VS Code
+   - Markdown All in One
+   - YAML
+
+## What's Included
+
+The settings file configures:
+- Format on save
+- Tab size matching `.editorconfig`
+- Trailing whitespace trimming
+- File exclusions for `node_modules`, build output
+- Search exclusions for session logs
+
+## Tips
+
+- VS Code's Copilot extension also reads `AGENTS.md` — no additional setup needed
+- Use the integrated terminal to run commands from `RUNBOOK.md`
+- Pin `TASKS.md` as a tab for quick reference
+
+## Troubleshooting
+
+| Issue | Solution |
+|-------|----------|
+| Settings not applied | Ensure `.vscode/` is at project root |
+| Formatter conflicts | Set `editor.defaultFormatter` to your preferred formatter |
+| Files not excluded | Check `files.exclude` patterns in settings.json |