OpenClaw Signals: Forcing Strict Schemas, Agent Registry Fixes, and Restart Races

Pull Request

Agent Space Accessor Pattern (PR #328)

Refactors how the agent registry is accessed, replacing direct calls with the agent_space.get() pattern. If you're hooking into the startup sequence or writing custom instances, this deduplicates state and prevents race conditions. Check if your private forks override AgentRegistry or if they pass the checks.

Breaking Change

Strict Typing and Orchestrator Migration (PRs #326 & #327)

Three separate PRs landed that tighten strict validated args on the ResponseFormatter, TypeScript migration of the orchestrator, and the AsyncIO EventLoop warning. They are enforcing stricter typing across the runtime to catch errors at compile time rather than runtime. If you are lagging behind on your type definitions, your CI is going to catch fire on the next sync.

Bug Report

Intermittent Context Loss on Agent Restart (Issue #1559)

A high-priority issue flagged by the community regarding intermittent context loss during agent restarts. The root cause appears to be a race condition in the new Redis state adapter, causing sandbox corruption if the restart triggers before the save signal is acknowledged. If you're running high-availability clusters, pin your Redis version and watch this thread closely until the patch lands.

Enhancement

Stricter Tool Schema Enforcement (Issue #1560)

The tool schema enforcement is getting stricter, moving from a warn to a throw state for malformed definitions. This is a good tactical shift for builders to clean up technical debt; error logs are getting excessively bloated with silent failures that just degrade performance. Treat this as a forcing function to fix your config files before upgrading.