Skip to content

Security: automatis-tools/agents-can-communicate

Security

SECURITY.md

Security policy

This page is for reporting a vulnerability. For system boundaries and threat reasoning, read the security model.

Report privately

Report privately via GitHub security advisories.

Please do not open a public issue first.

Include: what you ran, what happened, what you expected, and the client and version. A reproduction against a throwaway ACC_DATA_HOME is ideal.

Check the scope

In scope Out of scope
Peer text escaping its quoted block An attacker who already has write access to your data home
Any path escaping the managed root A model choosing to obey persuasive peer text
Uninstall deleting files ACC did not write Vulnerabilities in Codex, Claude Code, Gemini, Grok, or Kimi themselves
Unexpected ACC runtime/state or installer artifacts in a repository (excluding user-requested acc.workspace.json) Denial of service by a trusted peer
Session impersonation across MCP

Reasoning behind each: docs/SECURITY_MODEL.md.

Target a supported version

Pre-1.0: the latest release only.

Expect this response

Acknowledge, reproduce, and tell you whether it is in scope. If it is, the fix ships with a test in tests/security/ so it cannot come back quietly.

There aren't any published security advisories