Skip to content

Security: incordai/Insi

Security

SECURITY.md

Security

Please do not open a public issue for a vulnerability.

Report it privately with a GitHub Security Advisory on this repository (Security → Report a vulnerability). Include the version you run (insi --version), your platform, the command, and enough detail to reproduce it. You will get an answer, and credit in the fix if you want it.

What counts

Insi runs a coding agent on tasks, on issues anyone can write, and opens pull requests from what it produces. The most serious reports are the ones where one of its boundaries breaks:

  • The hidden check reaches the worker. A task's hidden tests, or its reference commit, readable by the agent before or during its attempt. This breaks every measurement Insi makes.
  • Untrusted text acts. An unlabelled issue worked by insi issue, or text from an issue, a connector or a CI log that makes Insi or its worker do something it should only have read.
  • A secret leaves the machine. A credential or personal data in a branch that insi issue or insi improve --pr pushed despite the scan.
  • The sandbox does not hold. Setup, the worker or the check reaching past the [sandbox] command that should contain them.
  • The MCP server or the hook writes. insi mcp and insi hook only read; any input that makes either write a lesson or run a command is a bug.

What to expect from Insi itself

Insi runs the worker command you configure, on your machine, with your permissions, unless you name a [sandbox]. It reads, holds and sends no GitHub token: gh holds the login. See the README for each command's limits.

Supported versions

Fixes are made on main and in the next release.

There aren't any published security advisories