Skip to content

SECURITY.md

Latest commit

 

History

History
66 lines (48 loc) · 1.95 KB

SECURITY.md

File metadata and controls

66 lines (48 loc) · 1.95 KB

Security

Threat model (summary)

  • User input may contain sensitive data (credentials, PII, keys)
  • Third-party plugins may be untrusted
  • Risks: data exfiltration (network), DoS (infinite loop), XSS in UI, supply-chain compromise

Mitigations

Execution isolation

Operations run inside a Web Worker:

  • No DOM access
  • Network APIs (fetch, XHR, WebSocket) overridden to throw at runtime
  • AbortSignal + hard worker termination for timeouts/DoS

CSP (production hosting)

Set these HTTP headers on your server:

Content-Security-Policy:
  default-src 'none';
  script-src 'self' 'wasm-unsafe-eval';
  style-src 'self';
  img-src 'self' data:;
  worker-src 'self';
  connect-src 'none';
  base-uri 'none';
  object-src 'none';
  frame-ancestors 'none'

Recommended companion headers for the static workbench runtime:

  • X-Content-Type-Options: nosniff
  • X-Frame-Options: DENY
  • Referrer-Policy: no-referrer
  • Permissions-Policy: accelerometer=(), camera=(), microphone=(), payment=()
  • Cross-Origin-Opener-Policy: same-origin
  • Cross-Origin-Resource-Policy: same-origin

The shipped container image enforces these headers in docker/nginx/default.conf, and pnpm docker:test verifies them during image smoke validation.

'wasm-unsafe-eval' is intentionally enabled so browser-safe hashing fallbacks based on WebAssembly can run when WebCrypto.subtle is unavailable.

Supply chain

  • pnpm v10: use allow-builds allowlist for trusted native dependencies
  • Commit pnpm-lock.yaml to the repository
  • Dependabot for automated dependency updates
  • CodeQL for static analysis (JavaScript/TypeScript)

No eval

  • eval() and new Function() are forbidden in production code
  • ESLint rule no-eval enforced
  • Dynamic import() from user-controlled strings is forbidden

Reporting vulnerabilities

Open a GitHub issue with label security or contact maintainers privately. Do not publish vulnerability details before a fix is available.