If you believe you've found a security issue in autopilot — such as a prompt-injection vector, a path-traversal in the JSONL writer/reader, an unsafe shell construction in a baked prompt, or any other credential / data-exposure risk — please report it privately rather than opening a public issue.

Use GitHub's private vulnerability reporting form to file a confidential security advisory. We'll triage and respond as soon as we can. Public-issue reports for credible vulnerabilities will be redacted on request.

When reporting, please include:

• A clear description of the vulnerability and its impact (what an attacker can do, what privileges/access they need to start with).
• Reproduction steps or a proof-of-concept that demonstrates the issue.
• The version / commit SHA you observed it on.
• Any suggested mitigations or patches you've prototyped.

This is a small, single-maintainer project with a low release cadence. Only the latest tagged release receives security updates. If a fix is applied, it ships in the next patch release on top of main; older tags are not back-patched.

If you need to keep using an older version, please cherry-pick the fix yourself or pin to the next tagged release that includes it.

In-scope:

• Every runtime module shipped under packages/tui/src/ and packages/tui/bin/.
• Anything baked into prompt templates that could be exploited by a malicious external input (e.g. a poisoned issue body fed into --grow-project).

Out-of-scope:

• Vulnerabilities in the Copilot CLI itself, Node.js, gh, or the Ink/React/Yoga/Commander dependency tree. Please report those to the respective projects.
• Issues that require an attacker to already have local code-execution on the user's machine (we assume the local environment is trusted).