Ai

Claude Mem v12.4.2 Improves Worker Recovery and Hook Stability

Claude Mem v12.4.2 Improves Worker Recovery and Hook Stability

Claude Mem v12.4.2 is a stability-focused release that addresses two ship-blocking issues discovered during triage, alongside several smaller fixes aimed at improving worker recovery, prompt hygiene, and cross-platform hook behavior. The update is especially important for teams relying on autonomous agent workflows, as it prevents poisoned sessions from trapping queued work and stops internal protocol payloads from being stored as user prompts.

What Changed

One of the biggest fixes in v12.4.2 improves worker reliability when the Claude SDK hits a context overflow. Previously, if the SDK returned a Prompt is too long error, the recovery flow could keep resuming the same broken session, causing the worker to loop until a restart guard eventually gave up. The new behavior clears session.memorySessionId and sets session.forceInit = true before throwing, ensuring the next crash-recovery spawn starts a completely fresh SDK session.

The release also prevents autonomous protocol messages wrapped in <task-notification> from being mistakenly stored in user_prompts. A new isInternalProtocolPayload() predicate in src/utils/tag-stripping.ts now filters these internal payloads at both the hook layer in session-init.ts and the worker boundary in SessionRoutes.ts. Importantly, this deny-list is conservative and does not interfere with <command-name> or <command-message> tags used for legitimate slash commands.

Several triage fixes round out the release. Issue #2092 updates the worker-service.cjs build banner to be CommonJS-safe by removing reliance on import.meta.url, allowing node -c validation to pass again. Issue #2100 corrects an apparent timeout typo by reducing the PreToolUse Read hook timeout from 2000 seconds to 60. Issue #2131 adds "shell": "bash" to every hook in plugin/hooks/hooks.json, improving compatibility for Claude Code on Windows by routing through Git Bash.

Why It Matters

This version matters because it focuses on operational resilience rather than surface-level features. The context overflow recovery fix directly addresses a failure mode that could strand large backlogs of pending messages on a poisoned session, making autonomous workers more dependable in production environments.

The prompt storage cleanup is also significant for data quality and observability. By keeping internal protocol payloads out of user prompt history, Claude Mem preserves cleaner records for debugging, analytics, and memory-related workflows without disrupting legitimate command traffic.

Together, the remaining fixes improve build correctness, reduce the risk of runaway hook waits, and strengthen Windows compatibility. For teams running Claude Mem in mixed environments or depending on unattended agent execution, v12.4.2 is a meaningful maintenance release centered on reliability and correctness.

Official Source: https://github.com/thedotmack/claude-mem/releases/tag/v12.4.2