Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/neuron-core/neuron-ai/agents-mdgit clone --depth 1 https://github.com/neuron-core/neuron-aiWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01450 | $0.01450 |
| Opus 5 | $0.00725 | $0.00725 |
| Sonnet 5 | $0.00290 | $0.00290 |
| Haiku 4.5 | $0.00145 | $0.00145 |
Grade A, and why
neuron-ai AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 139 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Neuron AI - Project Overview
This repository contains Neuron AI, a PHP framework for building agentic applications.
The Mental Model: Neuron Is a Composable Workflow
Before writing any Neuron code, internalize this — it shapes every correct decision:
-
Workflow is the foundation. Everything else is built on it. A Workflow is an event-driven graph of nodes that you can compose freely to create any agentic behaviour.
-
A catalog of components, wired by interfaces. Neuron ships a large collection of standalone, interchangeable building blocks — AI providers, the messaging layer, chat history, tools & toolkits, structured output, RAG, vector stores, data loaders, MCP, observability. They are connected by small, focused interfaces, so each can be used on its own or swapped for your own implementation.
-
Agentis an inspiration, not a cage. TheAgentclass is itself a Workflow: it composes the components above (chat/stream/structured nodes, tools, history, middleware) into a ready-to-use agentic entity. Study it to see how a Workflow is assembled. Then choose your path:- Extend
Agentwhen its defaults fit your use case (most common starting point). - Customize
Agentby swapping nodes or adding middleware when you need targeted changes without rebuilding from scratch. - Compose your own Workflow from the component catalog when you need bespoke control flow —
AgentandRAGcan even be used as nodes inside it.
- Extend
You never fight the framework: the architecture you use on day one is the same one running your most complex system later.
Development Commands
composer test # Run tests (PHPUnit)
composer format # Fix code style (PHP CS Fixer)
composer analyse # Static analysis (PHPStan level 5)
Individual tests: vendor/bin/phpunit tests/AgentTest.php or --filter testMethodName
Architecture
Layered Foundation:
RAG ─────► extends Agent ─────► built on Workflow
│
Tools ◄───────────────────────┤
Providers ◄───────────────────┤
Chat ◄────────────────────────┴──► shared across all
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 139 lines · 1,450 tokens per session scan A 9adfbe2f739f
neuron-ai AGENTS.md is an instructions file published in the GitHub repository neuron-core/neuron-ai (2,081 stars, last pushed yesterday), licensed MIT. It adds 1,450 tokens to every session, about $0.0072 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
code-cli AGENTS.md
Instructions for autohandai/code-cli, covering agents.md, project overview, current repository architecture, src/core/agent runtime split (current) and general layout guidance for contributions.
dive CLAUDE.md
Instructions for deepnoodle-ai/dive, covering claude.md, project overview, commands, architecture and core types.
reactive-agents AGENTS.md
Instructions for tylerjrbuell/reactive-agents, covering agents.md, quick start checklist, development commands, virtual environment setup and do not manually activate the virtual environment.
mini-swe-agent copilot-instructions.md
Copilot instructions for SWE-agent/mini-swe-agent, covering style guide, test style, bad and good.
plano CLAUDE.md
Instructions for katanemo/plano, covering claude.md, build & test commands, rust — wasm plugins (must target wasm32-wasip1), rust — brightstaff binary (native target) and rust — tests, format, lint.
Browser4 CLAUDE.md
Instructions for platonai/Browser4, covering browser4 — project context for claude, architecture, key dispatch chain (cli → browser), batch commands and e2e test structure.