feat: refactor agent integrations into a modular system and remove legacy integration directories

Author: fajarhide

CHANGELOG.md

@@ -14,6 +14,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - **Expanded Tool Registry**: Added granular `cargo` subcommand support and new tool categories for Database, Mobile, Cloud, and CI/CD toolchains with accurate distiller routing.
 
 ### Improved
+- **OpenClaw Portability**: The OpenClaw integration natively fetches plugin files directly from the public GitHub repository, allowing successful 1-click installation without requiring a full local git repository clone.
 - **Robust RegEx Generation (`omni learn`)**: Fixed a critical bug where auto-learned numeric patterns used literal `#` instead of functional `\d+` in generated TOML filters. Now delegates TOML string escaping to the `toml` crate for correctness.
 - **Enhanced Verify Report (`omni learn --verify`)**: Results are now grouped by source (Built-in vs. User), with clear per-category pass/fail counts and actionable tips when user-learned filters fail.
 - **Auto-Clear Learn Queue**: `omni learn --apply` now automatically clears `~/.omni/learn_queue.jsonl` after successful application, preventing stale data from polluting subsequent `--discover` runs.
@@ -23,6 +24,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - **Filter Loading**: Made `match_command` optional in `FilterConfig`, gracefully skipping filters with empty or missing patterns instead of crashing.
 
 ### Fixed
+- **Learned Filters Concurrency (`learned.toml`)**: Replaced seconds-based timestamp resolution with `timestamp_micros()` for auto-generated filters to prevent fatal TOML duplication parse errors during high-frequency concurrent learning (fixes the infinite `doctor --fix` `.bak` failure loop).
 - **Test Regression (`test_claude_code_stdout_format`)**: Resolved a persistent CI failure caused by state contamination from user-learned filters leaking into the test environment.
 - **Stats UX Hint**: Added `--all-commands` usage hint to `omni stats` when showing truncated top-10 results.
 

Cargo.lock

@@ -154,7 +154,7 @@ By default, `omni init --claude` automatically hooks into **Claude Code**. Howev
 
 1. **VS Code & Continue.dev**: Use our MCP context provider (`integrations/continue-dev/`).
 2. **OpenCode & Codex CLI**: Built-in wrappers automatically pipe command output to OMNI.
-3. **Generic Webhook (Antigravity IDE etc)**: Run `omni serve --port=7891` to launch an ultra-lightweight local HTTP server for agents to drop payloads into OMNI.
+3. **Antigravity IDE**: OMNI natively supports IDEs like Antigravity via MCP over stdio. Just import the plugin manifest (`integrations/antigravity/antigravity.plugin.json`).
 
 **Multi-Agent Tuning (`~/.omni/config.toml`)**
 Different agents have different pain points. Keep VS Code chat clean, whilst letting OpenCode read more data. Tune them individually:

README.md

@@ -898,10 +898,20 @@ omni init              # Interactive Menu for any AI Agent
 
 # Or bypass the menu for a specific agent
 omni init --claude     # Full Setup for Claude Code (Hooks + MCP)
-omni init --vscode     # Setup for VS Code Continue.dev
+omni init --cursor     # Setup for Cursor AI
+omni init --zed        # Setup for Zed Editor
+omni init --cline      # Setup for Cline
+omni init --roo        # Setup for Roo Code
+omni init --copilot    # Setup for GitHub Copilot CLI
+omni init --gemini     # Setup for Gemini CLI
 omni init --opencode   # Setup for OpenCode plugin
 omni init --codex      # Setup for Codex CLI
+omni init --openclaw   # Setup for OpenClaw plugin
+omni init --antigravity# Setup for Antigravity IDE
+
+# Claude-specific utilities
 omni init --mcp        # Setup Claude MCP Server only
+omni init --hook       # Setup Claude hooks only
 omni init --status     # Check Claude installation status
 omni init --uninstall  # Remove all OMNI components from Claude
 ```
@@ -1178,7 +1188,7 @@ OMNI is designed to be the "Intelligence Layer" for multiple agent frameworks.
 You can use OMNI natively with OpenClaw by installing the official skill from ClawHub:
 - **ClawHub**: [OMNI Semantic Signal Engine](https://clawhub.ai/fajarhide/omni-signal-engine)
 - **Install Command**: `clawhub install omni-signal-engine`
-- **Manual Install**: `openclaw plugins install ./integrations/openclaw`
+- **Automatic Setup**: OpenClaw integration is completely automatic. Run `omni doctor --fix` to fetch and install the latest plugin directly from the OMNI public repository.
 
 ---