extra AGENTS.md

Repository instructions for AI coding agents working on extra-org/extra. They describe the project's purpose, architecture, rules, folder structure, commands, and other expectations for making changes.

In plain words
What is it for?
Use it when an AI agent needs to understand, validate, generate, run, serve, or modify this YAML-based platform for building AI agent systems.
Why use it?
They give an agent the project context it needs before editing code. This reduces the risk of changes that conflict with the repository's design or development process.

Instructions file for CodexOpenCode

Install

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.

agentmods
npx agentmods add instructions/extra-org/extra/agents-md
Clone the repo
git clone --depth 1 https://github.com/extra-org/extra

Made for: Codex, OpenCode.

Per session 4,028 This file is loaded in full into every session.
When invoked 4,028 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.04028 $0.04028
Opus 5 $0.02014 $0.02014
Sonnet 5 $0.00806 $0.00806
Haiku 4.5 $0.00403 $0.00403

Measured 2d ago against content hash 989ae47478a4, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

extra 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 2d 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.

AGENTS.md · 358 lines

How it starts

The opening of the file, as written. The whole thing — 358 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md

This file is the operating manual for every AI coding agent working in this repository. Read it fully before making any change. If a task ever conflicts with this file, this file wins — stop and ask for clarification.


1. Project mission

This repository is a declarative platform for building AI agent systems.

Developers describe an agent system in YAML. The platform validates that YAML, compiles it into a typed internal graph, and runs it through a long-lived runtime that renders prompt files, calls resolver/tool plugins and MCP servers, exposes an API, and produces execution traces.

The repository is in active development. The YAML validator, compiler, runtime engine (orchestrators run as supervisor agents), resolver plugin system, tool plugin loading, MCP client (local and remote servers, including authenticated MCP via hooks), the runtime hooks system (11 lifecycle points — see docs/RUNTIME_HOOKS.md), per-run execution-limit guardrails (see docs/EXECUTION_LIMITS.md), prompt rendering, and the CLI (validate, inspect, generate, run, serve, chat) are implemented. Model access supports both Anthropic and Amazon Bedrock. Two HTTP API layers exist: a thin agent_engine API (/invoke, /stream), started by agentctl serve (default port 8090) — stateless, no persistence, no web client — and agent_manager — a conversation lifecycle service built on top of it with process-local storage by default, opt-in SQLite/Postgres persistence, SSE streaming, and the official React web client, started by the separate agent-manager console script (default port 8100). agentctl chat is a separate, ephemeral developer console that persists nothing (see docs/ARCHITECTURE.md §14 for detail). Basic observability (structured logging plus a Langfuse callback provider) is wired in. A Dockerfile/entrypoint.sh provide a basic container image.

The access plugin is wired into child filtering, but the request-context gate that should populate real identity/permissions into it is not yet implemented — access filtering currently runs against an empty context, so protected nodes are not actually enforced today. Treat this as an open security gap, not a finished feature. See docs/ROADMAP.md for per-phase status. Agents implement the product task-by-task using the files in tasks/, though the package layout has evolved beyond what the original numbered tasks describe (see §4).

Read the full file on GitHub · 358 lines

Changes

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.

  1. 2d ago First seen · 358 lines · 4,028 tokens per session scan A 989ae47478a4

Subscribe to this mod's changes

extra AGENTS.md is an instructions file published in the GitHub repository extra-org/extra (108 stars, last pushed 4d ago), licensed MIT. It adds 4,028 tokens to every session, about $0.0201 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.