Borrowing it
Nothing to install: this file belongs to provos/ironcurtain. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/provos/ironcurtain/master/.github/instructions/sandbox.instructions.mdgit clone --depth 1 https://github.com/provos/ironcurtainWrote 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/instructions/provos/ironcurtain/sandbox)<a href="https://agentmods.dev/instructions/provos/ironcurtain/sandbox"><img src="https://agentmods.dev/badge/instructions/provos/ironcurtain/sandbox/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/instructions/provos/ironcurtain/sandbox"><img src="https://agentmods.dev/badge/instructions/provos/ironcurtain/sandbox.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 | $0.00450 | $0.00450 |
| Opus 5.5 | $0.00180 | $0.00180 |
| Sonnet 5.5 | $0.00090 | $0.00090 |
| Haiku 4.5 | $0.00045 | $0.00045 |
Grade A, and why
ironcurtain sandbox.instructions.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 1mo 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 — 19 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Sandbox Review Rules
The sandbox registers a custom UTCP communication protocol (ironcurtain-protocol.ts) that routes tool calls in-process to the ToolCallCoordinator. It is the integration point between the agent's V8 isolate and the security boundary.
Mandatory Checks
- The sandbox MUST route tool calls through the
ToolCallCoordinatorvia the custom IronCurtain UTCP protocol (ironcurtain-protocol.ts). Never register real MCP servers directly with UTCP — that bypasses the entire policy engine. - MCP proxy subprocesses (
mcp-proxy-server.ts) are pure relay transports spawned byMCPClientManager. Security is enforced upstream in the coordinator, not in the proxy.PROXY_COMMANDandPROXY_ARGSstill exist for starting relay subprocesses. - Each backend MCP server gets its own relay subprocess via
SERVER_FILTERenv var. This ensures credential isolation (OAuth tokens, sandbox-runtime wrapping). - The per-sandbox unique manual name (e.g.,
tools_<uuid>) in the UTCP registration prevents cross-sandbox collision when multiple sandboxes run in the same process (daemon mode). Never hardcode the manual name. - The
Protocol.requesttimeout monkey-patch must execute beforeCodeModeUtcpClient.create(). Moving it later leaves client instances with the wrong timeout. - Tool functions inside the sandbox are synchronous. Do not add
awaitto sandbox tool calls. @utcp/mcpmust be imported (side-effect import) to register the MCP call template type. Removing this import breaks MCP functionality.- Strip internal fields at the sandbox boundary: when returning coordinator responses to the V8 isolate (or any untrusted execution context), explicitly filter to the public contract (
{ content, isError }). Do not rely on a specific transport's serialization behavior (e.g., the MCP SDK dropping unknown keys during stdio framing) to keep internal fields out — a new transport may not inherit that behavior. Internal-only fields like_policyDecisionmust never reach code running inside the isolate, where they leak policy-rule names and reasons to untrusted LLM-generated code.
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.
- 1mo ago First seen · 19 lines · 450 tokens per session scan A 92b71de1c341
ironcurtain sandbox.instructions.md is an instructions file published in the GitHub repository provos/ironcurtain (615 stars, last pushed yesterday), licensed Apache-2.0. It adds 450 tokens to every session, about $0.0018 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.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deno copilot-instructions.md
Copilot instructions for denoland/deno, covering deno development guide for github copilot, network access, high level overview, key directories and key files to understand first.