If you discover a security vulnerability in IntentFrame Agent Integrations, please report it responsibly. Do not open a public GitHub issue.
Email: intentframe@gmail.com
Please include:
- A description of the vulnerability
- Steps to reproduce it
- The potential impact
- Any suggested fix (optional)
We will acknowledge receipt within 48 hours and aim to provide an initial assessment within 7 days.
This repository integrates external agents with IntentFrame policy validation. Vulnerabilities in the following are in scope for this repo:
- Policy bypass through the Hermes plugin, adapter, or CLI wiring
- Governance or policy YAML handling that allows ungoverned tool execution
- Bridge authentication or authorization flaws in integration clients
- Install/uninstall scripts that expose secrets, corrupt user config, or leave privileged state behind unexpectedly
- Incorrect tool mapping that sends unvalidated side effects to Hermes handlers
Core IntentFrame runtime vulnerabilities (Guardian, Analysis Engine, native bundles) should be reported to the upstream IntentFrame project as well, since this repo depends on those packages at runtime.
Hermes Agent vulnerabilities should be reported upstream to Nous Research Hermes Agent.
- Vulnerabilities in third-party dependencies (report upstream)
- Issues that require physical access to the host machine
- Social engineering attacks against users
- Bugs in ungoverned Hermes tools that are intentionally not routed through
IntentFrame (documented in
docs/NATIVE_KIT_INTEGRATION.md)
We follow coordinated disclosure. We will work with you on a timeline and credit you in the advisory unless you prefer to remain anonymous.