This project takes security seriously. If you find a vulnerability — in the tooling, in the shared clients, or in any Live tool — please report it privately before disclosing publicly.

Email security@agentmindcloud.dev with:

• A clear description of the issue and its impact
• Steps or a minimal proof-of-concept to reproduce
• Affected tool(s) or package(s) and the version / commit SHA
• Your preferred contact for follow-up

Expected response: we aim to acknowledge within 72 hours and provide a remediation timeline within one week. Please do not open a public GitHub issue for security reports.

• The root monorepo and its build/CI tooling
• Every tool under tools/
• Every shared package under shared/
• Example code and templates shipped in this repo

Out of scope: third-party services this toolkit integrates with (X API, xAI Grok API, hosting platforms). Report issues in those to their vendors.

The project is pre-1.0. Only the latest 0.1.x minor line receives security fixes; older snapshots are archived.

Once we reach 1.0, this table will expand to cover the active major line plus the previous one.

Secrets (XAI_API_KEY, X_BEARER_TOKEN, any OAuth tokens) must follow these rules:

• Never committed. .env and .env.* are gitignored. The repo ships only .env.example placeholders with no real values.
• Server-side: load from environment variables at process start. Never log, never serialize into responses, never cache to disk.
• Client-side (browser): session-only. A user may paste a token into a tool's input field for that page load, but the tool must not persist it — no localStorage, no cookies, no query strings. The token lives in memory until the tab closes.
• Live tools with no backend (04, 05, 07, 09, 12, 18) do not accept tokens at all; they operate on public data or local heuristics only.
• Spec'd tools that need tokens proxy them through a serverless function or Node backend — never directly from the browser. See docs/cors.md for why.

See docs/tokens.md for setup, rotation, and per-tool requirements.

• Secret exfiltration, token leakage in logs/responses/errors
• XSS, CSRF, clickjacking, or prototype pollution in any Live tool
• Prompt-injection paths that coerce a Grok-backed tool into unintended actions
• Dependency issues with known CVEs in shipped package.json trees
• CI or supply-chain weaknesses (misconfigured actions, stale pinned versions with known issues)

Good-faith security research on your own fork or self-hosted copy is welcome. Please avoid testing against shared infrastructure you don't own, and never access data that isn't yours.

Part of the grok-install family. Apache 2.0.