Ait

v0.55.50 Security

This release includes 1 security fix for security teams reviewing exposed deployments.

Published 24d AI Agents & Assistants

View tool

✓ No known CVEs patched

Read the diff → Tool health → What is this tool? →

This release patches 1 known CVE

Topics

agent-harness agent-isolation agentic ai-agents ai-coding ai-tools

+14 more

aider claude-code code-provenance codex coding-assistant cursor developer-tools gemini-cli git git-worktree llm-agents python vcs worktree

Affected surfaces

auth rbac

ReleasePort's take

Light signal

editorial:auto 13d

Ait v0.55.50 hardens multi-agent branch landing with ref locks and stale-base checks to prevent unsafe automated commits. The release includes control plane documentation, embeds next_action in agent context, and fixes adversarial reviewer API key inheritance.

Why it matters: Teams using multi-agent branch automation should deploy this hardening to prevent unsafe landing. Test the adversarial reviewer fix and review new control plane documentation before deployment.

Summary

AI summary

Hardened multi-agent control plane prevents unsafe landings by using ref locks and stale‑base checks.

Changes in this release

Type	Severity	Summary	CVE
Security	Medium	Hardens same-target branch landing with ref locks and stale-base checks Hardens same-target branch landing with ref locks and stale-base checks Source: llm_adapter@2026-05-21 Confidence: low	—
Feature	Medium	Adds multi-agent control plane guide and acceptance coverage Adds multi-agent control plane guide and acceptance coverage Source: llm_adapter@2026-05-21 Confidence: high	—
Feature	Medium	Embeds next_action in primary agent context commands Embeds next_action in primary agent context commands Source: llm_adapter@2026-05-21 Confidence: high	—
Bugfix	Medium	Fixes adversarial reviewer API key inheritance with local claude -p Fixes adversarial reviewer API key inheritance with local claude -p Source: llm_adapter@2026-05-21 Confidence: high	—

Full changelog

Release 0.55.50 adds the multi-agent control plane guide and acceptance coverage, embeds next_action in primary agent context commands, hardens same-target branch landing with ref locks and stale-base checks, and fixes the built-in Claude Code adversarial reviewer to use local claude -p without inheriting ANTHROPIC_API_KEY.