Skip to content

Security: leavemagic-cyber/agent-aftercare

SECURITY.md

Security policy

Supported version

Agent Aftercare is pre-1.0 alpha software. Only the latest tagged release is supported with security fixes.

Threat model

The primary risks are deleting the wrong data, racing an active writer, restoring corrupted bytes, following a symlink outside a known root, and overwriting new user data.

The v0.1 boundary is intentionally narrow:

  • only known Codex, Claude Code, and Gemini CLI session-log roots;
  • only regular .jsonl files, never symlinks;
  • an exact process-name guard plus a hard recency gate;
  • capture, decompression, and source re-verification before removal;
  • atomic manifests with a crash-recoverable captured state;
  • no overwrite on restore;
  • no network, telemetry, elevated privilege, or background service.

Agent Aftercare does not protect against a malicious process running as the same operating-system user and modifying the quarantine store or executable. Keep the store under your own account and obtain binaries from this repository's release page.

Report a vulnerability

Please use GitHub's private Report a vulnerability flow under the repository's Security tab. Do not publish deletion, traversal, race, or restore integrity bugs as public issues before a fix is available.

Include the Agent Aftercare version, operating system, exact command, redacted directory shape, and whether the original data still exists. Never attach real session logs; they may contain private conversations.

There aren't any published security advisories