For anyone managing a personal knowledge base, ensuring its accuracy and structural integrity is an ongoing task. Information drifts, links break, and pages can become disconnected from the main body of work. This is where maintain comes in. It's an AI agent skill designed specifically for knowledge base diagnostics and correction, helping you keep your information organized and reliable.
maintain focuses strictly on the health of your knowledge base. It's not about generating new content or answering questions, but rather about ensuring the existing data is robust and well-connected. It's a tool for anyone who relies on a personal knowledge base and wants to avoid the common pitfalls of information decay.
Comprehensive Health Checks
maintain performs a suite of comprehensive health checks on your knowledge base. It scans for several structural problems that can degrade the utility of your stored information. For instance, it identifies orphaned pages – content that exists but isn't linked from anywhere else within your knowledge base, making it difficult to discover. It also diligently checks for broken links, ensuring that cross-references and external citations remain valid.
Beyond just links and orphans, maintain also flags stale information. This is content that hasn't been updated in a while and might be outdated, requiring review. Furthermore, it scans for citation gaps, highlighting areas where information could benefit from clearer sourcing or connections to other relevant data points. The goal is to provide a clear picture of your knowledge base's current state across these critical dimensions.
Remediation Paths and Scoring
Once problems are identified, maintain offers two distinct paths for remediation. The first is an autonomous mode, which you can activate with a command like doctor --remediate. In this mode, maintain will automatically attempt to fix detected issues. While running autonomously, it tracks the operational cost of its actions and aims to achieve a target health score, giving you an automated way to clean up your data with minimal intervention.
The second path is a manual, guided process. This is for users who prefer granular control over every decision. In this mode, maintain walks you through each identified dimension individually. For example, it will present all stale pages for your judgment, then move on to orphaned pages, dead links, backlinks, citations, filing rules, and tags. At each step, you retain full discretion to approve or reject suggested fixes.
Both paths offer clear reporting. After each health check, maintain provides before-and-after scores for each dimension it evaluated. This allows you to quantify the improvements made and track the ongoing health of your knowledge base over time. For example, you might see that your 'Broken Links' score improved from 60% to 95% after a maintenance run, clearly indicating progress.
Triggers and Advanced Modes
maintain can be triggered by specific commands related to knowledge base health. Common triggers include 'brain health', 'maintenance', 'orphan pages', and 'check backlinks'. These commands allow you to initiate a full scan or target specific problem areas as needed.
Beyond its core diagnostic and corrective functions, maintain also includes extra modes that enhance its utility. A 'dream cycle' mode synthesizes recent conversations and inputs, integrating them into your long-term memory structures within the knowledge base. This helps solidify new information and ensure it's properly filed and linked. Additionally, an 'autopilot' mode allows maintain to run continuous background maintenance loops, performing routine checks and minor corrections without direct user commands. This keeps your knowledge base in optimal condition passively.
FAQ
Q: What kind of problems does maintain address? A: It scans for structural issues like orphaned pages, broken links, stale information, and citation gaps within your personal knowledge base.
Q: Can maintain create new content or answer questions? A: No, maintain is strictly diagnostic and corrective. It focuses on the integrity of existing data, not content generation or synthesis.
Q: How does the autonomous remediation mode work?
A: Using a command like doctor --remediate, it auto-fixes issues while tracking the operational cost and working towards a target health score for your knowledge base.
Keeping a knowledge base healthy means proactively addressing its structural integrity. maintain provides the tools to do just that, ensuring your information remains accessible and reliable.
