This release includes 1 security fix for security teams reviewing exposed deployments.
Published 28d
Developer Productivity
✓ No known CVEs patched
This release patches 1 known CVE
Affected surfaces
auth
breaking_upgrade
Summary
AI summaryCodex project‑scope trust bootstrap now keeps AIRG‑managed tool approvals local to the trusted project .codex layer, enhancing security.
Full changelog
[2.3.1] - 2026-05-06
Changed
- Promoted the
2.3.1development train to release and aligned package, status, hook, and documentation version markers. - Telemetry architecture moved to an hourly scheduler that runs generator/uploader workers in parallel.
- Telemetry state now tracks
telemetry.last_payload_generated_dateandtelemetry.last_payload_uploaded_atin policy. - Codex integration is now scope-aware across MCP apply/remove, hardening, and posture detection.
- Codex hardening now uses dedicated scoped
rules/airg.rulespolicy-mirror files instead of managing Codexdefault.rules.
Added
- Telemetry outbox persistence in the runtime data directory (
<state_dir>/telemetry/telemetry-YYYY-MM-DD.json). - Telemetry service status and restart API endpoints:
GET /telemetry/service-statusPOST /telemetry/service-restart
- Advanced Policy telemetry UI now includes:
- service status modal (
status+last runfor generator/uploader) - warning banner for stale generator/failing uploader
- restart button.
- service status modal (
- Codex project-scope trust bootstrap in
~/.codex/config.tomlvia[projects."<workspace>"].trust_level = "trusted"with explicit GUI confirmation. - Codex project-only AIRG tool approval entries in
<workspace>/.codex/config.tomlfor[mcp_servers.ai-runtime-guard.tools.*].
Fixed
- Telemetry no longer depends on UTC day-rollover wake logic only; hourly scheduler runs continue while backend is active and stand down when no action is needed.
- Upload retries now persist naturally through outbox files instead of dropping failed sends.
- Fixed
write_file/edit_filetool crashes caused byNameError: name 'lower' is not definedincheck_path_policyextension checks. - Reports -> Log table now enforces fixed-width column truncation for long event/rule/command/path values so oversized text no longer expands row/cell boundaries.
- Reports -> Log now exposes consistent
Show moreactions for trimmedEvent,Matched Rule, andCommand / Pathvalues via modal expansion. - Codex text rewrite/apply flows are now idempotent across scoped
config.toml,AGENTS.md, andrules/airg.rules. - Settings -> Agents no longer shows false
Pending Changesfor hardening state when posture already matches the applied baseline.
Security
- Codex project scope now keeps AIRG-managed tool approvals local to the trusted project
.codexlayer instead of writing them to user~/.codex/config.toml.
Tests
- Added regression coverage for:
- telemetry generator/uploader scheduler behavior
- blocked-extension write/edit regression after path-policy fix
- Codex project-trust apply/remove behavior
- Codex project-scope hardening targets and idempotency
- Codex project-scope posture detection
- Codex project-only tool approval sections.
Security Fixes
- Codex project scope now confines AIRG‑managed tool approvals to the trusted project `.codex/config.toml` layer instead of writing them to user `~/.codex/config.toml`, preventing cross‑project approval leakage.
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
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.
Related context
Related tools
Beta — feedback welcome: [email protected]