Skip to content

Security: mission69b/t2000

Security

SECURITY.md

Security Policy

Supported Versions

Version Supported
8.x (latest major)
< 8

@t2000/{sdk,cli,mcp,id} release in lockstep — only the latest major receives security fixes.

Reporting a Vulnerability

Do NOT open a public GitHub issue for security vulnerabilities.

GitHub Security Advisories (preferred): Report a vulnerability

Include: a description, steps to reproduce, potential impact, and a suggested fix if you have one.

Response timeline

  • Acknowledgment: within 48 hours
  • Initial assessment: within 5 business days
  • Fix timeline: critical issues within 7 days, others within 30 days

Scope

  • @t2000/sdk — key handling, transaction building, gasless send / swap / pay, spend-limit enforcement, receipt verification (verifyReceipt)
  • @t2000/cli — input validation, wallet file handling (~/.t2000, 0600)
  • @t2000/mcp — DEPRECATED (stdio retired); the MCP surface is hosted Passport Connect (audric/apps/mcp)
  • @t2000/idagent_id::registry transaction builders
  • Move contracts (contracts/) — agent_id::registry (ownership / kill-switch authorization) and confidential_anchor (receipt-anchor integrity)
  • packages/serve — merchant-side x402: challenge issuance, settle verification, upstream API-key isolation
  • Websites (apps/docs) — XSS, injection

Out of scope

  • The retired @t2000/engine@4.x on npm (frozen legacy consumer app only)
  • Social engineering; DoS via rate limiting (implemented)
  • Vulnerabilities in third-party dependencies (report upstream)

Security Measures

  • Non-custodial keys — Ed25519, Bech32 JSON at 0600; the private key never leaves the user's machine. (No PIN/passphrase layer by design — the security boundary is the filesystem ACL; see ARCHITECTURE.md § Wallet + keys.)
  • Default-on spending limits — per-tx + daily caps enforced inside the SDK, gating CLI and MCP writes.
  • Payment verification — challenge-bound signed payments, structurally verified before settlement; the handler runs BEFORE money moves, so a failing seller never charges the buyer (no-charge-on-failure).
  • Confidential inference — an Audric product (api.audric.ai): fail-closed GPU-TEE attestation, TEE-signed receipts, Sui-anchored hashes.
  • Automated scanning — CodeQL + dependency audit in GitHub Actions.

Audit Status

Last audit: March 2026 (automated full-stack review) Status: findings remediated on a rolling basis — contact the maintainers for the report.

This is beta software. Use at your own risk.

There aren't any published security advisories