Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/miru-zero/zero-brainnpx agentmods add skills/miru-zero/zero-brain/sandbox-escape-techniquesWrote 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.
[](https://agentmods.dev/skills/miru-zero/zero-brain/sandbox-escape-techniques)<a href="https://agentmods.dev/skills/miru-zero/zero-brain/sandbox-escape-techniques"><img src="https://agentmods.dev/badge/skills/miru-zero/zero-brain/sandbox-escape-techniques/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.
<a href="https://agentmods.dev/skills/miru-zero/zero-brain/sandbox-escape-techniques"><img src="https://agentmods.dev/badge/skills/miru-zero/zero-brain/sandbox-escape-techniques.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00049 | $0.02467 |
| Opus 5 | $0.00024 | $0.01234 |
| Sonnet 5 | $0.00010 | $0.00493 |
| Haiku 4.5 | $0.00005 | $0.00247 |
Grade A, and why
sandbox-escape-techniques 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 6d 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.
This is a copy
94% identical to sandbox-escape-techniques — 4 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.
How it starts
The opening of the file, as written. The whole thing — 251 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SKILL: Sandbox Escape — Expert Attack Playbook
AI LOAD INSTRUCTION: Expert sandbox escape techniques across Python, Lua, seccomp, chroot, Docker/container, and browser sandbox contexts. Covers CTF pyjail patterns, seccomp architecture confusion, chroot fd leaks, namespace escape, and Mojo IPC abuse. Distilled from ctf-wiki sandbox sections and real-world container escapes. Base models often miss the distinction between sandbox types and apply wrong escape techniques.
0. RELATED ROUTING
- browser-exploitation-v8 — V8 exploitation for renderer RCE before browser sandbox escape
- container-escape-techniques — Docker/container specific escape techniques
- kernel-exploitation — kernel exploit for container/namespace escape
- linux-privilege-escalation — post-escape privilege escalation
Advanced References
- PYTHON_SANDBOX_ESCAPE.md — Full pyjail methodology:
__builtins__recovery, keyword bypass, AST bypass, pickle escape - SECCOMP_BYPASS.md — Architecture confusion, io_uring bypass, ptrace bypass, allowed syscall chaining
1. SANDBOX TYPE IDENTIFICATION
| Sandbox Type | Indicators | Typical Context |
|---|---|---|
| Python sandbox (pyjail) | Limited builtins, filtered keywords, exec/eval available |
CTF, online judges, Jupyter |
| Lua sandbox | No os, io modules; restricted metatables |
Game scripting, config |
| seccomp | syscall filtering, prctl(PR_SET_SECCOMP) |
CTF pwn, container hardening |
| chroot | Changed root filesystem, limited /proc access |
Legacy isolation |
| Docker/container | Namespaces, cgroups, reduced capabilities | Cloud, microservices |
| Browser (renderer) | OS-level sandbox (seccomp-bpf + namespaces on Linux) | Chrome, Firefox |
| Namespace isolation | PID/mount/network/user namespace | Container runtimes |
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 6d ago First seen · 251 lines · 49 tokens per session scan E 9632ced72534
sandbox-escape-techniques is a skill published in the GitHub repository miru-zero/zero-brain (0 stars, last pushed 22d ago), licensed MIT. It adds 49 tokens to every session and 2,467 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. It is 94% identical to sandbox-escape-techniques, differing in 4 lines, and is treated as a copy.
Other skills, from other repositories
deploy-docker-compose
Run the Omnigent server as a Docker compose stack (server + Postgres) on any Docker host — your laptop, a VPS, EC2 by hand, or as the base layer of any container-platform deploy. Invoke when the user wants to build the image, bring up the compose stack, debug the stack on a host they already have, or extend the stack…
compute-env-setup
Set up a reproducible Feynman compute environment for research jobs. Use when a task needs Python/R packages, GPU libraries, containers, Modal, SSH, caches, or managed model runtime setup.
securing-kubernetes-on-cloud
This skill covers hardening managed Kubernetes clusters on EKS, AKS, and GKE by implementing Pod Security Standards, network policies, workload identity, RBAC scoping, image admission controls, and runtime security monitoring. It addresses cloud-specific security features including IRSA for EKS, Workload Identity for…
detecting-privilege-escalation-in-kubernetes-pods
Detect and prevent privilege escalation in Kubernetes pods by monitoring security contexts, capabilities, and syscall patterns with Falco and OPA policies.
implementing-rbac-hardening-for-kubernetes
Harden Kubernetes Role-Based Access Control by implementing least-privilege policies, auditing role bindings, eliminating cluster-admin sprawl, and integrating external identity providers.
docker-socket-mount
Docker / containerd socket mounted into a container → host RCE. Common in CI runners, GitOps controllers (ArgoCD, Flux), and 'Docker-in-Docker' setups. Single-command escape via docker run --rm --privileged -v /:/host alpine chroot /host.