How this repo works
Build, deploy, verify path
The Configuration Control Plane for AI Coding Agents: Claude Code, Cursor, Cline, Copilot, Codex, Antigravity & Grok
One source of truth (AGENTS.md), one sync engine, every agent aligned — plus a curated index of the operational layer that keeps autonomous coding agents safe, fast, and consistent.
Use this repo when you need one canonical instruction layer for Claude Code, Cursor, Cline, Copilot, Codex, Antigravity, Grok, or a mixed agent fleet.
git clone https://github.com/frankxai/agentic-ops-hub.git
cd agentic-ops-hub
node scripts/sync-agent-rules.mjs --checkStart with:
| I want to... | Start with |
|---|---|
| Understand the canonical rule source | AGENTS.md |
| Fan rules out to agent-specific files | scripts/sync-agent-rules.mjs |
| Copy the baseline into another repo | templates/AGENTS.md |
| Understand the operating ecosystem | ECOSYSTEM.md |
| See MCP and protection strategy | docs/MCP-STRATEGY.md and docs/PROTECTION-LAYERS.md |
This is a library/configuration repo, not a deployable app. The fastest health check is the generated-rule drift check above.
This repo is also the canonical map of the agentic-income ecosystem — the L0–L7 layer model, every repo's purpose and status, the MCP own-vs-adopt strategy, the founder/queen/worker agent stack, the protection layers for humans + agents + wealth, and the red/blue charter.
ECOSYSTEM.md— the canonical map: layers, repos, money path.docs/MCP-STRATEGY.md— which MCP servers we own vs adopt (incl. AP2 / x402 / ACP).docs/AGENT-STACK.md— founder agent · stream queens · workers · escalation contract.docs/AGENT-OPERATING-SYSTEM-STACK.md— how AGENTS.md, rules, hooks, skills, MCP, Hermes, OpenClaw, DeepAgents, Claude Code, and Codex map into the broader agent OS stack.docs/PROTECTION-LAYERS.md— defense-in-depth for humans, agents, and wealth.docs/RED-BLUE-CHARTER.md— what red attacks, what blue defends, the cadence.
This repository exists to draw — and own — an exact distinction:
| AIOps (the old term) | Agentic Ops (this repo) | |
|---|---|---|
| What it operates | Enterprise IT infrastructure: logs, metrics, alerts | Autonomous AI coding agents: Claude Code, Cursor, Cline, Codex, Grok |
| Core mechanism | ML models detecting anomalies in telemetry | Rules, guardrails, hooks, skills, and MCP servers governing agent behavior |
| Who runs it | SRE / platform teams | Developers and creators running multi-agent fleets |
| Failure mode prevented | Outages, alert fatigue | Drifted instructions, rogue edits, clobbered worktrees, token waste |
| Unit of config | Dashboards, runbooks | AGENTS.md, SKILL.md, hooks, rule files |
AIOps uses machine learning to monitor infrastructure. Agentic Ops is the developer-centric discipline of configuring, orchestrating, guarding, and aligning the AI agents that write your code. Different layer, different operator, different failure modes. This repository is the home of Agentic Ops.
Agent instruction files fragmented into a dozen formats. The 2026 resolution is layered:
AGENTS.md ← universal base (the agents.md standard, plain markdown)
├─ CLAUDE.md ← thin shim: @AGENTS.md + Claude-only additions
├─ .cursor/rules/*.mdc ← generated (frontmatter, alwaysApply / glob-scoped)
├─ .clinerules/*.md ← generated (directory format)
├─ .github/copilot-instructions.md ← generated
└─ .claude/skills/coding-guardrails/SKILL.md ← generated (auto-activating skill)
You edit one file. The sync engine generates the rest with tamper-evident headers. CI verifies nothing drifted (--check).
.cursorrules(single file) is deprecated — Cursor still reads it, but new projects should use.cursor/rules/*.mdc. The sync script emits the legacy file only with--legacy.
# 1. Copy the canonical source + sync engine into your project
cp templates/AGENTS.md /path/to/project/AGENTS.md
cp templates/CLAUDE.md /path/to/project/CLAUDE.md
mkdir -p /path/to/project/scripts
cp scripts/sync-agent-rules.mjs /path/to/project/scripts/
# 2. Fan out to all agents
cd /path/to/project && node scripts/sync-agent-rules.mjs
# 3. (CI) Fail the build if any generated rule file drifted
node scripts/sync-agent-rules.mjs --checkTemplates ship the Top Thinkers Guardrails (Karpathy, Feynman, Ousterhout, Hickey, Torvalds, Beck) and a Multi-Agent Coordination Protocol for running several harnesses against one repo without clobbering each other.
Agentic Ops is a stack. Each repo owns one layer — agentic-ops-hub is the config control plane that aligns them:
| Layer | Repo | Owns |
|---|---|---|
| Config control plane | agentic-ops-hub (you are here) | Rule source-of-truth, cross-agent sync, the Agentic Ops index |
| Capability system | agentic-creator-os | 90+ skills, 65+ commands, 38 agents — what agents can do |
| Lifecycle enforcement | claude-code-hooks | Quality gates, circuit breakers, audit trails — what agents may do |
| Integration health | mcp-doctor | Diagnose/optimize MCP servers — what agents connect to |
| Machine health | peak-performance | System auditing for agent-heavy machines — what agents run on |
| Memory substrate | second-brain-os · Starlight-Intelligence-System | Persistent knowledge + the SIP protocol — what agents remember |
| Prompt layer | prompt-engine · prompt-library | Evaluated, red-teamed prompts — what agents are told |
| Domain expertise | claude-skills-library | Deep research-backed domain skills — what agents know |
Rule of thumb: capabilities live in ACOS, enforcement lives in hooks, configuration alignment lives here. If a file tells multiple agents how to behave in a repo, this hub owns its lifecycle.
The best external repositories, toolkits, and skills for operating AI coding agents:
- forrestchang/andrej-karpathy-skills — The original viral Karpathy rules for Claude Code.
- vercel-labs/agent-skills — Vercel's official prompt library for coding agents.
- wsimmonds/claude-nextjs-skills — Evaluators and NextJS testing patterns.
- agentsmd/agents.md — The AGENTS.md standard: one plain-markdown file, every agent reads it.
- modelcontextprotocol — MCP, the integration standard for agent ↔ tool connectivity.
- affaan-m/ECC — Everything Claude Code. Commands, skills, and security audits via
AgentShield. - smithery-ai/mcp-servers — Index of MCP integrations.
- giuseppe-trisciuoglio/developer-kit — Operational scripting and helper toolkits.
- sickn33/antigravity-awesome-skills — 500+ skills for Antigravity, Grok, and Codex environments.
/templates—AGENTS.md(canonical source),CLAUDE.md(shim), legacy.cursorrules/.clinerules, ACOSSKILL.md./scripts—sync-agent-rules.mjs: fan-out +--checkCI verification +--legacycompat./docs—layering.md: what goes in which file, and why;CODING_AGENTS.md: selection matrix and inventory of coding agents.
MIT License. Free to use, fork, and distribute. Contribution PRs are welcome!