| name | agbridge-parallel-ide-agent |
| description | Control and interact with the Antigravity Bridge Daemon (`agbridge`) for asynchronous codebase analysis and parallel agent execution. This skill provides instructions on how to use the CLI and manage workspace sessions. |
Agbridge Daemon Skill
This skill allows you to control the Antigravity Bridge Daemon (agbridge). By utilizing the daemon, you can trigger deep codebase analysis while continuing to work, or manage parallel analysis sessions across multiple workspaces.
🛡️ Constraints & Permissions
- Project Files: READ-ONLY. You are strictly FORBIDDEN from editing, creating, or deleting source code files.
- Web Access: ALLOWED & ENCOURAGED.
- Execution via Daemon: Use the
agbridge CLI to interact with the LLM orchestrator asynchronously.
🚀 How to use agbridge Command (Daemon Integration)
The agbridge command allows you to control the background daemon and interact with the AI agents. Here is the full list of available commands and how to apply them to the analyzer workflow.
1. Requesting Analysis (ask)
Use ask to send the analyzer prompts to an agent. Describe the scope of analysis clearly.
agbridge ask "Please act as an analyzer. Analyze the current codebase architecture. Do not modify files."
agbridge ask --new "Analyze the dependencies in package.json and find security vulnerabilities."
agbridge ask --session <session_id> "Expand on the previous analysis regarding the database layer."
agbridge ask --workspace /path/to/project "Analyze the routing logic of this project."
agbridge ask --no-queue "Quickly read the README and summarize the core features."
2. Managing Analysis Sessions (session)
Use session to manage past or parallel analysis threads.
agbridge session list
agbridge session connect <session_id>
agbridge session show <session_id>
3. Monitoring the Daemon (status & info)
Verify that the daemon is correctly connected to the target workspace before sending large analysis jobs.
agbridge status
agbridge status --workspace /path/to/project
agbridge info
agbridge info --refresh
4. Advanced Debugging (debug)
If the AI agent is struggling to read UI state or requires low-level DOM inspection.
agbridge debug tree
agbridge debug tree --depth 20
For detailed human-readable guides, please refer to: resources/agbridge_guide_en.md and resources/agbridge_guide_ko.md
5. Tracking Conversation History (.ag-sessions)
If you want to view, track, or manage the local conversation history of your daemon interactions, look for the .ag-sessions folder inside your target workspace.
- Directory Location:
<TARGET_WORKSPACE_PATH>/.ag-sessions/
- Internal Structure:
history_<session_id>.json: Contains the raw conversation log (user inputs and AI responses) for a specific background session.
active_session.txt: Tracks which session ID is currently active for the workspace.
Tracing these files is highly recommended when you need to audit the agent's analysis context or verify the JSON data generated during deep codebase scans.