skill/isol8/SKILL.md
Securely execute untrusted Python, Node.js, Bun, Deno, and Bash code in sandboxed Docker containers.
npx skillsauth add Illusion47586/isol8 isol8Install this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Isol8 is a secure execution engine for running untrusted code inside Docker containers with strict resource limits, network controls, and output sanitization. Use this skill when you need to execute code, scripts, or system commands in a safe, isolated environment.
For full documentation, see the isol8 docs. This file is a quick-reference for AI agents — it covers the most common operations and links to detailed docs for everything else.
| Command | Purpose | Full Docs |
|:--------|:--------|:----------|
| isol8 run [file] | Execute code in an isolated container | CLI: run |
| isol8 setup | Build Docker images, optionally bake in packages | CLI: setup |
| isol8 cleanup | Remove orphaned isol8 containers (optionally images with --images) | CLI: cleanup |
| isol8 serve | Start HTTP server for remote execution (downloads binary on first use) | CLI: serve |
| isol8 config | Display resolved configuration | CLI: config |
isol8 run)--eval flag (inline code, defaults to python runtime)--runtime)python runtime)Extension mapping: .py → python, .js/.mjs/.cjs → node, .ts → bun, .mts → deno, .sh → bash
Remote source input: use --url, --github, or --gist (requires remoteCode.enabled=true in config). Add --hash for SHA-256 verification.
isol8 run)| Flag | Default | Description |
|:-----|:--------|:------------|
| -e, --eval <code> | — | Execute inline code |
| -r, --runtime <name> | auto-detect | Force: python, node, bun, deno, bash |
| --no-stream | false | Disable real-time output streaming |
| --persistent | false | Keep container alive between runs |
| --persist | false | Keep container after execution for debugging |
| --debug | false | Enable internal debug logging |
| --install <package> | — | Install package before execution (repeatable). If --net is omitted, CLI auto-uses filtered and allowlists default runtime registries |
| --url <url> | — | Fetch source code from URL |
| --github <owner/repo/ref/path> | — | GitHub shorthand for raw source |
| --gist <gistId/file.ext> | — | Gist shorthand for raw source |
| --hash <sha256> | — | Verify SHA-256 hash for fetched code |
| --allow-insecure-code-url | false | Allow insecure http:// code URLs for this request |
| --net <mode> | none | Network: none, host, filtered (explicit value is never overridden) |
| --timeout <ms> | 30000 | Execution timeout |
| --memory <limit> | 512m | Memory limit |
| --secret <KEY=VALUE> | — | Secret env var, value masked in output (repeatable) |
| --stdin <data> | — | Pipe data to stdin |
For the complete flag reference (20 flags total), see CLI: run.
# Python inline
isol8 run -e "print('Hello!')" --runtime python
# Run a file (runtime auto-detected)
isol8 run script.py
# With package installation
isol8 run -e "import numpy; print(numpy.__version__)" --runtime python --install numpy
# Pipe via stdin
echo "console.log(42)" | isol8 run --runtime node
# Secrets (masked as *** in output)
isol8 run -e "import os; print(os.environ['KEY'])" --runtime python --secret KEY=sk-1234
# Remote execution
isol8 run script.py --host http://server:3000 --key my-api-key
# Cleanup orphaned containers
isol8 cleanup # Interactive (prompts for confirmation)
isol8 cleanup --force # Skip confirmation
isol8 cleanup --images --force # Also remove isol8 images
For full library documentation, see Library Overview.
import { DockerIsol8 } from "isol8";
const isol8 = new DockerIsol8({
mode: "ephemeral", // or "persistent"
network: "none", // or "host" or "filtered"
memoryLimit: "512m",
cpuLimit: 1.0,
timeoutMs: 30000,
secrets: {}, // values masked in output
persist: false, // keep container after execution for debugging
debug: false, // enable internal debug logging
});
await isol8.start();
const result = await isol8.execute({
codeUrl: "https://raw.githubusercontent.com/user/repo/main/script.py",
codeHash: "<sha256>",
runtime: "python",
installPackages: ["numpy"], // optional
});
console.log(result.stdout); // captured output
console.log(result.exitCode); // 0 = success
console.log(result.durationMs);
await isol8.stop();
Full options reference: Execution Options
import { RemoteIsol8 } from "isol8";
const isol8 = new RemoteIsol8(
{ host: "http://localhost:3000", apiKey: "secret" },
{ network: "none" }
);
await isol8.start();
const result = await isol8.execute({ code: "print(1)", runtime: "python" });
await isol8.stop();
for await (const event of isol8.executeStream({
code: 'for i in range(5): print(i)',
runtime: "python",
})) {
if (event.type === "stdout") process.stdout.write(event.data);
if (event.type === "exit") console.log("Exit code:", event.data);
}
Full streaming docs: Streaming
await isol8.putFile("/sandbox/data.csv", "col1,col2\n1,2");
const buf = await isol8.getFile("/sandbox/output.txt");
Full file I/O docs: File I/O
Full endpoint reference: Server Endpoints
| Method | Path | Auth | Description |
|:-------|:-----|:-----|:------------|
| GET | /health | No | Health check |
| POST | /execute | Yes | Execute code, return result |
| POST | /execute/stream | Yes | Execute code, SSE stream |
| POST | /file | Yes | Upload file (base64) |
| GET | /file | Yes | Download file (base64) |
| DELETE | /session/:id | Yes | Destroy persistent session |
| POST | /cleanup | Yes | Trigger remote cleanup for sessions/containers (images optional) |
Config is loaded from (first found): ./isol8.config.json or ~/.isol8/config.json. Partial configs are deep-merged with defaults.
Full configuration reference: Configuration
| Layer | Default |
|:------|:--------|
| Filesystem | Read-only root, /sandbox tmpfs 512MB (exec allowed), /tmp tmpfs 256MB (noexec) |
| User isolation | Non-root sandbox user (uid 100), processes killed between pool reuses |
| Processes | PID limit 64, no-new-privileges |
| Resources | 1 CPU, 512MB memory, 30s timeout |
| Network | Disabled (none), iptables enforcement in filtered mode |
| Output | Truncated at 1MB, secrets masked |
| Seccomp | strict default profile (blocks mount, swap, ptrace, etc.); standalone server binaries use embedded fallback when profile files are missing |
Container Filesystem:
/sandbox (512MB): Working directory, packages installed here, execution allowed for .so files/tmp (256MB): Temporary files, no execution allowed for securityFull security model: Security
isol8 setup to check.--timeout. Process is killed on timeout.--memory.--sandbox-size (default 512MB) or --tmp-size (default 256MB).--sandbox-size large enough for installation (512MB+ recommended)..ts files running with Bun instead of Deno: .ts defaults to Bun. Use --runtime deno or .mts extension.isol8 serve --update to force a fresh download. Use isol8 serve --debug to see detailed server logs. For listen port selection, precedence is --port > ISOL8_PORT > PORT > 3000; if the port is busy, serve can prompt for another port or auto-pick an available one.strict/custom mode, execution fails if profile loading fails. Verify security.customProfilePath for custom mode.development
Maintainer-only workflow for handling GitHub Secret Scanning alerts on OpenClaw. Use when Codex needs to triage, redact, clean up, and resolve secret leakage found in issue comments, issue bodies, PR comments, or other GitHub content.
development
Maintainer workflow for OpenClaw releases, prereleases, changelog release notes, and publish validation. Use when Codex needs to prepare or verify stable or beta release steps, align version naming, assemble release notes, check release auth requirements, or validate publish-time commands and artifacts.
development
Run, watch, debug, and extend OpenClaw QA testing with qa-lab and qa-channel. Use when Codex needs to execute the repo-backed QA suite, inspect live QA artifacts, debug failing scenarios, add new QA scenarios, or explain the OpenClaw QA workflow. Prefer the live OpenAI lane with regular openai/gpt-5.4 in fast mode; do not use gpt-5.4-pro or gpt-5.4-mini unless the user explicitly overrides that policy.
development
End-to-end Parallels smoke, upgrade, and rerun workflow for OpenClaw across macOS, Windows, and Linux guests. Use when Codex needs to run, rerun, debug, or interpret VM-based install, onboarding, gateway smoke tests, latest-release-to-main upgrade checks, fresh snapshot retests, or optional Discord roundtrip verification under Parallels.