Skip to content

great_cto

v2.43.0 Feature

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

Topics

agentic-coding claude-code-plugin claude-code-skills claude-code-subagents code-review cto
+2 more
multi-agent sdlc

Affected surfaces

auth rbac

ReleasePort's take

Moderate signal
editorial:auto 5d

Version v2.43.0 now requires a human checkpoint before any irreversible action can be executed and adds an accountable owner field to autopilot configurations.

Why it matters: Runtime blocks autonomous execution of irreversible actions without prior human checkpoint; enforce with validateFlow() for compliance.

Summary

AI summary

Enforced runtime invariant makes any irreversible action require a prior human checkpoint and assigns an accountable owner.

Changes in this release

Breaking High

Runtime now blocks autonomous execution of irreversible actions without prior human checkpoint.

Runtime now blocks autonomous execution of irreversible actions without prior human checkpoint.

Source: llm_adapter@2026-06-07

Confidence: high

Breaking High

Adds reversible and blastRadius fields to every flow step.

Adds reversible and blastRadius fields to every flow step.

Source: llm_adapter@2026-06-07

Confidence: low

Feature Medium

Introduces validateFlow() and flow-runner.mjs <v> --validate to enforce irreversible ⟹ preceded by a gate.

Introduces validateFlow() and flow-runner.mjs <v> --validate to enforce irreversible ⟹ preceded by a gate.

Source: llm_adapter@2026-06-07

Confidence: high

Feature Low

Adds accountable owner field to each autopilot for traceability.

Adds accountable owner field to each autopilot for traceability.

Source: llm_adapter@2026-06-07

Confidence: high

Bugfix Low

Adds 7 new tests, increasing flow-runner suite from 30 to 37 and total lib tests to 220.

Adds 7 new tests, increasing flow-runner suite from 30 to 37 and total lib tests to 220.

Source: llm_adapter@2026-06-07

Confidence: high

Full changelog

Inspired by Oleksandr Torlo's The Permission Was the Wound (2026): the danger of an agent isn't going rogue — it's doing exactly what it's permitted, irreversibly, at machine speed, with no human hesitation. This release makes the human↔autopilot boundary an enforced runtime invariant.

  • reversible + blastRadius on every flow step — read/draft/check vs money-move/claim-submission/e-signing/fleet-change/tax-filing.
  • The runtime refuses to execute an irreversible action autonomously — an irreversible step with no prior human checkpoint is blocked; a protected one runs only after the gate is signed (never auto-fired in a dry-run).
  • validateFlow() + flow-runner.mjs <v> --validate — enforces irreversible ⟹ preceded by a gate and every autopilot names an accountable owner. All six verticals ship green.
  • Accountable owner per autopilot — one named human answers for it, closing the confused-deputy / 'the AI did it' gap. Surfaced in the run trace and on each autopilot page.

Backward-compatible (new fields optional; all six flows already place irreversible actions after the human gate). 7 new tests — 30 in the flow-runner suite, 220 lib total.

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 great_cto

Get notified when new releases ship.

Sign up free

About great_cto

Engineering-management layer of 34 specialist AI agents covering the full SDLC (architect, PM, senior-dev, reviewer, QA, security, devops, L3-support + 18 archetype-specific reviewers) with auto-detected archetypes and compliance gates (PCI-DSS, HIPAA, FedRAMP, GDPR, EU AI Act). Runs in Claude Code, Cursor, Codex CLI, Aider, and Continue via AGENTS.md + MCP. MIT.

All releases →

Related context

Earlier breaking changes

  • v2.55.0 Hard-gates the Build board; redirects invite sessions to Operate for operators.
  • v2.32.0 Removed AgentShield scanner and its CLI commands.

Beta — feedback welcome: [email protected]