一键导入
team-bernie
This skill should be used when Antigravity needs to coordinate a team of intelligent agents working in parallel on the same project.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
This skill should be used when Antigravity needs to coordinate a team of intelligent agents working in parallel on the same project.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | team-bernie |
| description | This skill should be used when Antigravity needs to coordinate a team of intelligent agents working in parallel on the same project. |
| category | orchestration |
| risk | medium |
| tags | [collaboration, agents, multi-agent] |
| date_added | 2026-04-03 |
This skill allows Antigravity to coordinate a team of intelligent agents working in parallel on the same project, replicating the functionality of "Agent Teams".
The team uses a hidden folder in the project root to communicate:
.antigravity/team/tasks.json -> Master list of tasks, statuses, and dependencies..antigravity/team/mailbox/ -> Individual messages (.msg)..antigravity/team/broadcast.msg -> Global messages for the entire team..antigravity/team/locks/ -> Semaphores to prevent simultaneous file editing.Before making significant changes, each agent must send an Action Plan to Bernie's mailbox. The agent remains in READ_ONLY or PLANNING mode until Bernie responds with an APPROVED message.
broadcast.msg to give new directives to the entire team simultaneously.Tasks in tasks.json can have a list of dependencies. An AI should not claim a task if its dependencies are not in the COMPLETED state.
.lock exists in .antigravity/team/locks/.This script automates task management and communication. Save it as team_manager.py.
import json
import os
import sys
TEAM_DIR = ".antigravity/team"
def init_team():
"""Initializes the team infrastructure."""
os.makedirs(f"{TEAM_DIR}/mailbox", exist_ok=True)
os.makedirs(f"{TEAM_DIR}/locks", exist_ok=True)
tasks_path = f"{TEAM_DIR}/tasks.json"
if not os.path.exists(tasks_path):
with open(tasks_path, 'w') as f:
json.dump({"tasks": [], "members": []}, f, indent=2)
if not os.path.exists(f"{TEAM_DIR}/broadcast.msg"):
with open(f"{TEAM_DIR}/broadcast.msg", 'w') as f: f.write("")
print("✓ 'Bernie Team' infrastructure ready.")
def assign_task(title, assigned_to, deps=[]):
"""Assigns a new task with dependency support."""
path = f"{TEAM_DIR}/tasks.json"
with open(path, 'r+') as f:
data = json.load(f)
task = {
"id": len(data["tasks"]) + 1,
"title": title,
"status": "PENDING",
"plan_approved": False,
"assigned_to": assigned_to,
"dependencies": deps
}
data["tasks"].append(task)
f.seek(0)
json.dump(data, f, indent=2)
print(f"✓ Task {task['id']} ({title}) assigned to {assigned_to}.")
def broadcast(sender, text):
"""Sends a message to all team members."""
msg = {"de": sender, "tipo": "BROADCAST", "mensaje": text}
with open(f"{TEAM_DIR}/broadcast.msg", 'a') as f:
f.write(json.dumps(msg) + "\n")
print(f"✓ Global message sent by {sender}.")
def send_message(sender, receiver, text):
"""Sends a message to a specific agent's mailbox."""
msg = {"de": sender, "mensaje": text}
with open(f"{TEAM_DIR}/mailbox/{receiver}.msg", 'a') as f:
f.write(json.dumps(msg) + "\n")
print(f"✓ Message sent to {receiver}.")
if __name__ == "__main__":
if len(sys.argv) > 1:
cmd = sys.argv[1]
if cmd == "init": init_team()
iOS 26 Liquid Glass Design System — Dynamic glass materials for SwiftUI, UIKit, and WidgetKit with blur, reflection, and interactive morphing effects.
Conduct technical research on technologies and architecture. Use when the user says they would like to do or produce a technical research report
This skill should be used when the user types the /thatsallfolks trigger phrase to end the session, generate the handoff, and commit the changes.
This skill should be used when the user types the /borabill trigger phrase to start an Agentic Governance Framework session.
Centralized documentation of business rules, domain logic, and specific project requirements.
Learning repository that records successes, errors, and structured decisions to avoid repeating failures and feed the main-skill's memory system.