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 (incode --version), your platform, the agent and model, and enough detail to reproduce it. You will get an answer, and credit in the fix if you want it.

Incode runs every tool a model calls through its own guards, so the most serious reports are the ones where that promise breaks:

• a tool call that runs, writes or deletes something a guard should have refused, or that escapes the OS sandbox;
• text from a repository, a file, a web page or a tool's output that makes the agent take an action the user did not approve (prompt injection that crosses a permission boundary);
• a key, token or other secret written to a log, a transcript, the memory store or a commit;
• computer use acting outside what the user allowed.

Incode uses the agent CLIs you already have, run headless on your own subscriptions. A session can edit files and run commands, within the mode and permissions you chose. Read Safety and Security notes in the README for what each mode allows and what Incode will not do.

Fixes are made on main and in the next release.