Skip to content

Security: arthuz-th/AgentLocalCI

Security

SECURITY.md

Security policy

Supported versions

AgentLocalCI is pre-1.0. Security fixes are applied to the latest published release only.

Version Supported
latest 0.2.x beta Yes
0.1.x alpha and older snapshots No

Reporting a vulnerability

Do not open a public issue for a suspected credential exposure, cleanup bypass, host escape, dependency-proxy bypass, provenance failure, unsafe installer/profile mutation, or hook ownership bypass.

Use the repository's Security → Report a vulnerability private advisory form. Include:

  • affected commit or release;
  • host OS/architecture, PowerShell, Git, Docker Desktop/Engine versions;
  • exact command and minimal synthetic repository fixture;
  • expected versus observed boundary;
  • whether any host file, credential, network endpoint, Docker resource, hook, shell profile, or report was exposed;
  • a redacted proof of concept.

Do not include real secrets or private repository contents. Use synthetic canaries.

Security scope

High-priority reports include:

  • target code gaining host-file, Docker-socket, device, credential, or host-port access;
  • validation networking that is not none;
  • dependency traffic escaping the exact HTTPS allowlist;
  • cleanup deleting without exact owner/run/kind proof;
  • accepting a branch, tag, abbreviation, uppercase SHA, missing object, non-commit, or dirty tree through an easy command;
  • source history, remotes, host checkout, or private metadata entering validation;
  • installer or uninstall mutation outside owned paths/profile markers;
  • overwriting or removing a foreign Git hook;
  • secrets surviving report/log redaction or entering the public repository.

The current local user, host administrator, and Docker daemon operator are trusted. A malicious trusted host operator is outside the sandbox boundary.

There aren't any published security advisories