Skip to content

Security: FlowmemoryAI/FlowMemory

Security

SECURITY.md

Security Policy

FlowMemory is public research and local/test infrastructure. Do not treat this repository as production-ready, audited, mainnet-ready, or approved for uncapped value-bearing use.

Reporting Vulnerabilities

Use GitHub private vulnerability reporting / Security Advisories for suspected vulnerabilities. Do not open public issues containing exploitable details, private keys, RPC credentials, wallet mnemonics, API keys, webhook URLs, or user data.

Current Security Boundary

  • Heavy AI, model, memory, media, and artifact data stays off-chain.
  • On-chain contracts store compact roots, receipts, commitments, attestations, proofs, and intentionally bounded work state only.
  • Current public-agent, Agent Bonds, Rootflow, and Flow Memory paths are local/test or capped-pilot surfaces unless a document explicitly says otherwise.
  • Public test keys and deterministic fixtures may appear in tests and scripts. Real secrets must never be committed.
  • Base mainnet canary references are historical, bounded test evidence only; they are not a production launch.

Before Running Live Commands

Read:

  1. docs/CURRENT_STATE.md
  2. docs/PRODUCTION_READINESS_CHECKLIST.md
  3. docs/MARKETING_CLAIMS_GUARDRAILS.md
  4. the task-specific deployment or operations runbook

Any command requiring .env, RPC URLs, deployer keys, API keys, or wallet credentials must be run locally with uncommitted environment files.

There aren't any published security advisories