atft-autonomy

Coordinate Claude and Codex agents for ATFT-GAT-FAN stack maintenance.

Updated Aug 28, 2025
One-click install
npx skills add https://github.com/wer-inc/gogooku3 --skill atft-autonomy
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: atft-autonomy
Source: https://github.com/wer-inc/gogooku3/tree/main/claude/skills/atft-autonomy
Command: npx skills add https://github.com/wer-inc/gogooku3 --skill atft-autonomy

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires openai-codex-cli, and includes scripts (resource) components.

What problem does it solve?

Managing the complex, multi-faceted maintenance of the ATFT-GAT-FAN stack requires coordinating multiple AI agents (Claude and Codex) and maintaining shared context across diverse tasks. This Skill automates this orchestration, ensuring continuous system health.

Core Features & Use Cases

  • AI Agent Orchestration: Seamlessly coordinate Claude Code skills with OpenAI Codex autonomous workflows for end-to-end system maintenance.
  • Proactive Self-Healing: Schedule and execute daily autonomous maintenance loops, including quick fixes by Claude and deep optimizations by Codex.
  • Use Case: Set up a daily cron job to run a full autonomous maintenance loop, allowing Claude to handle quick fixes and Codex to perform deep optimization sweeps across dataset, training, and research modules, ensuring the system is always production-ready without manual intervention.

Quick Start

Example: Run full daily autonomy loop

./tools/claude-code.sh --no-check "Run proactive maintenance checklist" ./tools/codex.sh --max --exec "Perform deep optimization sweep across dataset, training, and research modules"

Frequently Asked Questions about atft-autonomy

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automate maintenance for the ATFT-GAT-FAN stack using AI agents?

Automate ATFT-GAT-FAN stack maintenance by orchestrating Claude and Codex agents to perform proactive maintenance across dataset, training, and quality domains. The Skill coordinates agent workflows, applies preflight validations, synchronizes shared context, and logs all actions to ensure continuous production readiness without manual intervention.

Can I schedule daily autonomous maintenance loops with Claude and Codex?

Yes. Schedule daily cron jobs to run full autonomy loops where Claude handles quick fixes and Codex performs deep optimization sweeps. The Skill manages agent coordination, regenerates configuration files when missing, updates documentation, and maintains audit logs across all maintenance operations.

What does multi-domain incident triage mean in AI agent orchestration?

Multi-domain incident triage coordinates Claude and Codex agents to diagnose and resolve issues across dataset, training, and quality domains simultaneously. The Skill verifies tool availability, synchronizes shared context between agents, and records resolutions to enable consistent, traceable incident handling across the entire stack.

Do I need to manually manage shared configuration when running autonomous agents?

No. The Skill automates shared-config management by regenerating .mcp.json when missing, updating AGENTS.md documentation, and synchronizing context across Claude and Codex workflows. This eliminates manual configuration drift and ensures all agents operate with consistent, current settings.

What preparation is required before running autonomous maintenance with this Skill?

Ensure openai-codex-cli is installed as a dependency and your ATFT-GAT-FAN environment is accessible. The Skill performs preflight validations to verify tool availability, regenerates missing configurations, and sets up audit logging—no additional manual setup needed.

How does the Skill handle agent coordination failures or incomplete maintenance?

The Skill implements preflight validations to detect issues before execution and maintains audit logs of all agent actions and outcomes. If coordination fails, logs reveal exactly which domain or agent encountered problems, enabling rapid diagnosis and rerun of specific maintenance tasks.