memory-curator

What This Skill Does

The memory-curator is a specialized utility designed to manage the cognitive load and long-term storage requirements of an AI agent. By transforming verbose, raw daily logs—often stretching into hundreds of lines—into compact, highly structured digests, this skill ensures that critical information remains accessible while keeping the context window clean. It focuses on distilling the day's activity into key sections: a high-level summary, actionable metrics, significant events, intellectual learnings, interpersonal connections, persistent questions, and tomorrow's objectives. This process is not merely about deletion; it is about indexing intelligence. By utilizing a hierarchical folder architecture (Index, Digests, Topics, and Raw Logs), the memory-curator allows the agent to maintain a clear trajectory of thought without becoming bogged down in granular, historical noise.

Installation

To integrate this skill into your environment, use the OpenClaw command-line interface: clawhub install openclaw/skills/skills/themiloway/memory-curator Once installed, you can trigger the generation of a digest skeleton for the current date using ./scripts/generate-digest.sh or specify a past date by passing it as an argument.

Use Cases

• Memory Architecture Maintenance: Building an index-first memory system where the agent quickly scans summaries before drilling into raw data.
• Token Optimization: Significantly reducing token overhead for long-term memory retrieval by prioritizing high-value summary data over repetitive, raw log files.
• Workflow Continuity: Preserving the 'state of mind' of an agent across sessions, ensuring that open questions and upcoming tasks are carried over from one day to the next.
• Automated Reporting: Serving as a bridge for end-of-day reporting, allowing the system to track progress and milestones through automated cron jobs.

Example Prompts

1. "Curator, please analyze today's raw logs and generate a summary digest, highlighting my interactions with the development team and any blockers encountered."
2. "Update my master index and include the latest daily digest from 2026-01-30. Ensure all persistent questions from that day are flagged for today's review."
3. "Summarize the last three days of raw logs into a single retrospective digest, focusing on the progress made in the Moltbook project."

Tips & Limitations

• Aggressive Compression: Remember that the goal is reconstruction, not transcription. If an event is trivial and documented elsewhere, omit it from the digest.
• Consistency: The effectiveness of this tool relies on your commitment to filling in the manual sections ('Summary', 'Learnings', 'Tomorrow') immediately after the skeleton is generated.
• Limitations: The tool performs text transformation and file management; it is not a semantic search engine on its own. It is most effective when paired with an agent that understands the directory structure specified in the architecture guide. Always maintain raw logs as a fallback, as the summary process is inherently lossy.