Skip to content

Security: Rayfts/AgentTrace

Security

SECURITY.md

Security Policy

AgentTrace records and replays developer-tool activity, so traces may contain source code, commands, paths, prompts, tool payloads, and other sensitive engineering data.

Reporting a vulnerability

Do not publish exploit details, credentials, sensitive traces, replay bypasses, redaction bypasses, or remote-access weaknesses in a public issue.

If GitHub private vulnerability reporting is enabled, use Security → Report a vulnerability. Otherwise, open a minimal non-sensitive issue requesting a private reporting channel.

Include the affected commit/version, operating system, minimal reproduction, impact, and whether the issue can expose trace contents, local files, credentials, execute unapproved commands, or bypass loopback/redaction protections.

Security boundaries

  • No automatic trace upload exists; storage is local-first.
  • Environment capture is deny-by-default and must not indiscriminately persist secrets.
  • Redaction is applied before normal persistence and again on read/export surfaces.
  • The local API binds to loopback by default; non-loopback binding requires explicit opt-in.
  • Replay is dry-run-first and requires explicit execution plus an exact recorded command/sequence allowlist.
  • Replay's temporary Git worktree protects repository state but is not an operating-system sandbox.
  • Raw payloads can still contain sensitive information; users should review traces before sharing them.

Changes to redaction, replay, API binding, artifact handling, or credential scrubbing should include focused regression tests and relevant documentation updates.

There aren't any published security advisories