Impasse sends artifact content to a third-party AI provider and supervises a subprocess. Please read docs/security-model.md for the trust boundaries (data boundary, prompt-injection, untrusted reviewer output, independence limits).

Please report suspected vulnerabilities privately — do not open a public issue for a security problem. Use GitHub's private vulnerability reporting on this repository (Security → Report a vulnerability), or contact the maintainer directly.

Include: what you found, how to reproduce it, and the impact. We'll acknowledge and work a fix before any public disclosure.

• Impasse does not scan for secrets and makes no guarantee that sensitive data is kept out of a review. Approving a send is the operator's responsibility; prefer allowlisting inputs.
• Reviewer output and reviewed artifacts are untrusted. Reports about the tool interpreting them as trusted content (prompt injection, escape-sequence rendering, executing cited paths) are in scope.
• Delegate mode is experimental and outside the trusted path; treat findings there accordingly.