Skip to content
Poligon

Security

Poligon sends hostile text at agents, so it is built to be a poor tool for attacking anyone else.

What we store

  • Your name, your email address and a password hash, or the FleetQ identifier if you sign in with FleetQ.
  • Teams, roles and invitations.
  • For each agent: its name, URL and the kind of authentication it needs. A secret you enter for the agent is stored encrypted, sealed so that only the runner can open it, and never shown again.
  • Runs: scenario results, episode traces and the planted texts. Runs are kept for 180 days, and a team owner can delete them earlier.

The tool services in a test are simulations. Nothing real is sent, paid or read.

Where Poligon will connect

  • Only to HTTPS addresses on the public internet. Every address a name resolves to is checked, and the connection goes to the address that was checked.
  • Never to private, loopback, link-local, carrier-grade NAT, multicast or Docker addresses, and never through a redirect.
  • Only after you have shown control of the agent's host with a file at /.well-known/poligon-verify.txt. See the agent protocol.
  • With limits per team: 2 runs at the same time, 20 runs a day and 200 episodes per run.

Accounts

  • Registration needs an invitation sent to your address. An account is never merged with another because the emails match; a FleetQ account is linked only from a signed-in session after you confirm your password.
  • Passwords need at least 12 characters and are checked against known breaches. Sign-in attempts are rate limited.

Signed reports

Each finished run is signed with an ed25519 key held by the Poligon runner. The public key is published at /.well-known/poligon-signing-key.pem. A shared report verifies the signature in the viewer's browser against that key.

Reporting a problem

If you find a security problem in Poligon, reply to the address that sent your invitation. Please do not run tests against agents you do not control.