Skip to content

This release adds 3 notable features for engineering teams evaluating rollout.

✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Affected surfaces

auth breaking_upgrade

Summary

AI summary

Added agent-scoped context to runtime, logs, approvals and explicit default blocked patterns for destructive wrappers.

Full changelog

[1.3.0] - 2026-03-03

Added

  • Connection-scoped identity/session context in runtime and logs:
    • agent_id and agent_session_id are now carried through tool execution and audit events.
    • reports filtering now includes session-level attribution.
  • Approval UX context improvements:
    • approvals now display agent-aware request context in the GUI while keeping full command details expandable.
  • Destructive wrapper policy coverage made explicit and transparent:
    • added default blocked command patterns for destructive wrapper forms (find -delete, find -exec rm, xargs rm, xargs -0 rm, do rm).
    • non-destructive find flows remain allowed by default.

Changed

  • Command safety behavior is now more policy-driven for destructive wrapper forms, with less hidden command-specific branching in runtime logic.
  • Default backup root behavior now resolves to user runtime state paths (<state_dir>/backups) for installed/runtime mode.
  • airg-doctor diagnostics now include resolved backup_root and warnings for unsafe backup-root placement (site-packages or project directory).

Fixed

  • Backup creation path fallback that could resolve under package directories in some installed-mode cases.
  • Backup gating consistency:
    • write_file and delete_file now honor audit.backup_enabled consistently with execute_command.
  • Documentation now tracks a known telemetry limitation where execute_command may undercount affected_paths_count for some shell-expanded/wrapper forms.

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Track jimmyracheta/AI-Runtime-Guard

Get notified when new releases ship.

Sign up free

About jimmyracheta/AI-Runtime-Guard

Runtime policy enforcement for AI agents - prevents accidental damage to your systems, unauthorized agent access and automates backup-before-write for any touched files.

All releases →

Related context

Beta — feedback welcome: [email protected]