Skip to content

Security: Samarth-23-eng/scope-intelligence

SECURITY.md

Security Policy

Supported versions

Version Supported
Latest release Yes
main Best effort
Older releases No

Supported deployment

Scope Intelligence is designed for a private, local workspace. Docker Compose binds the dashboard and API to 127.0.0.1. The current release does not include multi-user authentication and must not be exposed directly to the internet.

If you deploy behind a network boundary, use TLS, authentication, request limits, restrictive firewall rules, a reviewed reverse proxy, and independent backup and monitoring controls.

Reporting a vulnerability

Use GitHub's private vulnerability reporting for this repository. Do not open a public issue containing exploit details, credentials, private company data, or personal information.

Include:

  • affected version or commit;
  • component and deployment model;
  • minimal safe reproduction;
  • practical impact;
  • suggested mitigation, if known.

You should receive an acknowledgement within seven days. Timelines for a fix depend on severity, reproducibility, and maintainer availability.

Credential handling

  • Copy .env.example to .env; never commit .env.
  • Prefer short-lived, least-privilege, read-only credentials.
  • Protect LLM_MASTER_KEY and the Docker llm_secrets volume.
  • Never put credentials in investigation prompts, company records, URLs, screenshots, reports, fixtures, issues, or logs.
  • Revoke a credential immediately if it appears in a chat, terminal transcript, screenshot, commit, or build output.

Deleting a secret from the latest revision is insufficient when it remains in Git history. Revoke it, replace or rewrite the history, and run a full-history secret scan.

Scope

High-priority reports include authentication or authorization bypass, remote code execution, credential disclosure, cross-company data exposure, unsafe deserialization, server-side request forgery, prompt-driven external actions, and failures in secret encryption or redaction.

Automated scanning that degrades services, testing against third-party sources, and reports based solely on unsupported scanner output are out of scope.

There aren't any published security advisories