What is llm-jail AGENTS.md?

A set of project instructions for braiins/llm-jail, a Nix project that runs coding agents inside isolated QEMU virtual machines. It explains the project structure, build commands and end-to-end checks.

It helps work on the Nix configuration, run builds and checks, understand the virtual-machine architecture, and perform the documented end-to-end test.

It gives an agent the repository-specific rules needed to build, test and modify the project safely and consistently.

How braiins/llm-jail configures Codex and OpenCode

Reuse

Borrowing it

Nothing to install: this file belongs to braiins/llm-jail. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/braiins/llm-jail/main/AGENTS.md

Clone the repo

git clone --depth 1 https://github.com/braiins/llm-jail

Made for: Codex, OpenCode.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/instructions/braiins/llm-jail/agents-md/github.svg)](https://agentmods.dev/instructions/braiins/llm-jail/agents-md)

Your own site

<a href="https://agentmods.dev/instructions/braiins/llm-jail/agents-md"><img src="https://agentmods.dev/badge/instructions/braiins/llm-jail/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/instructions/braiins/llm-jail/agents-md"><img src="https://agentmods.dev/badge/instructions/braiins/llm-jail/agents-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 2,284 This file is loaded in full into every session.

When invoked 2,284 The same file — it is already loaded in full.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

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.

Model	Per session	Once invoked
Fable 5.1	$0.02284	$0.02284
Opus 5.5	$0.00914	$0.00914
Sonnet 5.5	$0.00457	$0.00457
Haiku 4.5	$0.00228	$0.00228

Measured 10d ago against content hash ea2d4089387f, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

llm-jail 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 10d 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 · 45 lines

How it starts

The opening of the file, as written. The whole thing — 45 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Build & Test

nix build .#claude              # Build the runner script
nix run .#claude -- --help      # Show CLI usage
nix flake check                 # Validate flake structure

End-to-end test (needs a terminal and a prior in-jail login - run nix run .#claude once and complete the login flow, which persists credentials in ~/.config/llm-jail/claude/default):

nix run .#claude -- --dangerous -- -p "Write hello to /workspace/hello.txt" --max-turns 3

After changes to guest NixOS config, rebuilds are fast (only systemd units regenerate). Changes to flake.nix inputs or writeShellApplication trigger full rebuilds.

Architecture

This is a Nix flake that runs coding agents inside QEMU microVMs with hardware-level isolation. No disk images - the guest boots on tmpfs with the host's /nix/store shared read-only via 9p (or virtiofs, see Key Constraints). An overlayfs covers /nix/store and /nix/var is bind-mounted from the same backing volume so build artifacts land on disk (not root tmpfs) when --store-disk is used.

Data flow: flake.nix iterates tools.nix, builds a NixOS guest system per tool, wraps each with lib/mkRunner.nix into a QEMU launcher script.

Host side (lib/mkRunner.nix): A writeShellApplication that parses CLI args at runtime, writes env vars and tool args to a temp dir, sets up 9p virtfs mounts (or one virtiofsd per share), optionally creates a store disk image, and launches QEMU with direct kernel boot. QEMU's stdout (the guest console stream) is not wired to the terminal directly but piped through pv -q -B 64M, which drains it eagerly into a 64MB buffer and writes to the tty as it becomes writable - QEMU's virtconsole silently drops guest output when the host fd can't absorb a burst (see Key Constraints), which corrupted TUI frames on busy terminals. On NixOS hosts, /run/current-system/sw and /etc/profiles/per-user/$USER are auto-mounted so host packages are available in the guest.

Read the full file on GitHub · 45 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.

10d ago Changed · +1 lines · +225 tokens per session ea2d4089387f
22d ago Changed · +1 lines · +242 tokens per session b84cbd704f66
2mo ago First seen · 43 lines · 1,817 tokens per session scan A f4057f779876

Subscribe to this mod's changes

llm-jail AGENTS.md is an instructions file published in the GitHub repository braiins/llm-jail (32 stars, last pushed yesterday), licensed MIT. It adds 2,284 tokens to every session, about $0.0091 per session on Opus 5.5. 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.

llm-jail AGENTS.md next to 6 instructions from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
llm-jail AGENTS.mdbraiins/llm-jail	Instructions file	32	yesterday	MIT	A	2,284	Codex, OpenCode
vscode buildNext.instructions.mdmicrosoft/vscode	Instructions file	195k	yesterday	MIT	A	9,719	GitHub Copilot
next.js AGENTS.mdvercel/next.js	Instructions file	143k	yesterday	MIT	A	7,461	Codex, OpenCode
langchain AGENTS.mdlangchain-ai/langchain	Instructions file	149k	yesterday	MIT	A	4,680	Codex, OpenCode
vscode oss-third-party-notices.instructions.mdmicrosoft/vscode	Instructions file	195k	yesterday	MIT	A	5,516	GitHub Copilot
vscode copilot-instructions.mdmicrosoft/vscode	Instructions file	195k	yesterday	MIT	A	3,093	GitHub Copilot
deepseek-harness AGENTS.mddeepseek-ai/deepseek-harness	Instructions file	274k	2d ago	MIT	A	4,313	Codex, OpenCode

llm-jail: Instructions file for Codex