memory-sync

What This Skill Does

Memory Sync is a specialized utility for the OpenClaw agent designed to maintain historical continuity across model swaps and session resets. By parsing raw JSONL session logs, it reconstructs and backfills your agent's memory files. The tool offers two distinct processing tiers: a high-speed simple extraction for rapid synchronization and an LLM-powered narrative mode that generates human-readable summaries. Crucially, Memory Sync includes an integrated sanitization layer to detect and redact secrets automatically during the processing pipeline, ensuring that sensitive information is not persisted in long-term memory logs.

Installation

To install, ensure you have Python 3.11 or higher installed. Navigate to your OpenClaw skills directory and use the clawhub command:

clawhub install openclaw/skills/skills/mpesavento/memory-sync

For manual local dependencies, run:

pip install click

Use Cases

• Model Transitions: When upgrading from one LLM version to another, use compare and backfill to ensure the new model has full context of past interactions.
• Daily Hygiene: Automate your memory maintenance by running a nightly cron job with the --summarize and --preserve flags for a perfectly curated, evolving knowledge base.
• Recovery: If an agent's context window clears or a memory file is corrupted, use backfill --all to rebuild the timeline from raw logs.
• Audit Compliance: Utilize the validate and stats commands to verify memory coverage and ensure important decisions are not lost.

Example Prompts

• "OpenClaw, run a memory sync for today's sessions and summarize the key technical decisions made."
• "Compare my current memory file with the latest logs and backfill any missing interactions from this week."
• "Perform an incremental memory backfill using LLM summarization while preserving my existing custom notes."

Tips & Limitations

• Preservation: Always use the --preserve flag if you have added manual annotations to your memory files; otherwise, the tool may overwrite your custom notes.
• LLM Cost: The --summarize mode triggers API calls. If you are on a restricted budget, default to standard extraction for large date ranges and reserve LLM summarization for daily incremental updates.
• Sanitization: While the tool features automated secret detection, always perform a manual audit of generated memory files containing highly sensitive PII or credentials.