Skip to content

Security: webcull/webcull-cli

Security

SECURITY.md

Security Policy

Reporting Issues

Report security issues privately to WebCull before public disclosure. Include a short description, affected command or endpoint, reproduction steps, and impact. Use GitHub private vulnerability reporting if it is enabled for this repository, or contact WebCull through the support channel listed on webcull.com.

Do not include live tokens, passwords, E2EE passphrases, or customer data in a report.

CLI Token Safety

The CLI and local MCP store tokens in OS credential storage with no plaintext fallback. Their authorizations are independent. Token values must not be printed, logged, committed, included in screenshots, or copied into bug reports.

The local config file may store only non-secret metadata such as expiry, account display data, and scopes.

E2EE Safety

The CLI accepts one E2EE passphrase piped directly from a user-managed keystore with explicit --e2ee-passphrase-stdin and account selection. Local MCP reads the selected account's OS keystore entry directly. Neither interface accepts E2EE secrets through command arguments, environment variables, configuration, logs, or AI tool arguments. See MCP setup for platform instructions.

The WebCull API must not receive passphrases, passphrase hashes, derived keys, decrypted check values, or decrypted bookmark fields. Decrypted results requested through MCP are shared with the connected AI application. Local decryption does not imply the AI application or its provider runs locally.

WebCull Account Access

The CLI connects to WebCull through the WebCull API and uses the permissions granted to the signed-in account. Keep your WebCull account, local machine, and credential storage protected.

Release Safety

Before publishing a release:

  • Run tests and syntax checks.
  • Inspect npm pack --dry-run.
  • Confirm package contents are limited to public CLI files.
  • Confirm no local config, credentials, build output, or unrelated project files are included.
  • Publish only from a clean trusted checkout with npm two-factor authentication enabled.

There aren't any published security advisories