Skip to content

Security: agenttrust-labs/agenttrust

Security

SECURITY.md

Security policy

Reporting

Use GitHub Private Vulnerability Reporting. Don't open a public issue. (security@agenttrust.tech is not yet configured for forwarding — fall back to the GitHub channel.)

Scope

In scope: the three Anchor programs (programs/policy-vault/, programs/trustgate/, programs/validation-registry/), the SDK (@agenttrust-sdk/trustgate), the MCP server (@agenttrust-sdk/mcp), the facilitator service (trustgate/server/), and the hosted endpoints at mcp.agenttrust.tech / api.agenttrust.tech / demo.agenttrust.tech.

Out of scope (please report upstream): Quantu (agent-registry-8004, atom-engine), Solana runtime, Anchor, MPL Core, and the four facilitators (Pay.sh / Dexter / atxp / MCPay) themselves — adapter-side bugs against those facilitators ARE in scope.

Response

Solo maintainer; best-effort triage. Acknowledgement target: 7 days. Standard 90-day coordinated-disclosure window from initial report; earlier disclosure if a fix lands sooner.

Documented footguns

docs/proofs/transfer-hook-atomicity.md is the kind of failure-class this policy already documents publicly (splitting gate_payment + transfer + emit_feedback corrupts state on Token-2022 + TransferHook mints). New issues in that style are exactly what this policy covers.

There aren't any published security advisories