Many of us accumulate vast amounts of digital information across various devices and cloud services over years. Sifting through this personal archive to find what's truly valuable – your own forgotten writing, unique ideas, or important relationship records – can be a significant undertaking. The new AI agent skill, archive-crawler, is designed to assist with this specific challenge. It intelligently scans your personal file collections, including local drives, Dropbox, and even Gmail takeouts. The primary goal of this tool is to surface valuable content you might have overlooked, while effectively filtering out the digital noise, such as system files and binary blobs, that often clutters personal archives. This skill empowers you to regain control over your digital past, making your own data more accessible and useful.
How the Skill Works: The Three Phases
The operational flow of this skill is structured into three distinct phases, each designed to progressively refine and organize your digital content.
The first phase involves an inventory of your file tree. During this initial scan, the skill maps out your designated file locations. It then ranks folders based on their likely relevance to personal intellectual property and records, distinguishing between directories likely containing meaningful content versus those filled with transient data or application files. This intelligent prioritization helps in focusing subsequent efforts on the most promising areas of your archive.
Following the inventory, the second phase focuses on presenting gold-filtered items for your review. This is where it truly acts as an assistant rather than an autonomous system. It surfaces what it identifies as valuable content – such as documents containing your original thoughts, drafts, or important communications – one item at a time. The system then captures your exact reactions to each item. This interactive review process is crucial; your feedback guides the tool's understanding of what constitutes 'valuable' for you personally, ensuring that the final output aligns with your subjective criteria. Imagine you've got years of email takeouts and old project folders spread across different hard drives. The skill might present a Markdown file you wrote years ago with brainstorming notes for a project you've since revisited, or an old email where you detailed a solution to a technical problem you're currently facing. Your review allows you to tag, categorize, or even discard these suggestions based on current relevance.
The third and final phase handles the ingestion of content into canonical directories. Based on your explicit review and specified filing rules, the tool will then organize the selected content. It uses predefined, yet customizable, directories such as originals/, personal/, and ideas/. This structured filing ensures that once content is identified and validated by you, it's placed in a logical and easily retrievable location, reducing future search overhead and bringing order to disparate files.
Designed for
Control and Supported Formats
Operating this skill is straightforward through specific triggers. You can initiate its functions using natural language commands such as 'crawl my archive', 'find gold in my archive', or 'mine my old files for'. These triggers prompt the tool to begin its scanning and review process in the designated paths.
A core design principle of archive-crawler is user control and data safety. It incorporates a critical safety gate: the skill strictly requires an explicit archive-crawler.scan_paths: configuration entry within your gbrain.yml file. This means the tool will never, under any circumstances, scan sensitive content like tax documents, medical records, or other confidential information unless you have specifically and intentionally listed those paths for scanning. This configuration requirement ensures that you maintain full control over which parts of your personal archive the skill can access, preventing uncontrolled data exposure.
Regarding file types, it supports a range of common document and email formats. The tool can process plain text files, HTML documents, Markdown files, .mbox email archives, Microsoft Word .docx files, .pst Outlook data files, and various compressed archives (e.g., .zip, .tar.gz). This broad support covers many typical personal data formats. However, it's important to understand its limitations: it is NOT designed for binary media processing (e.g., video or image analysis), automated classification without your explicit review and feedback, or bypassing any established privacy boundaries. The skill prioritizes a controlled, user-driven interaction over autonomous operations.
Practical
Archiving and Boundaries
The primary utility of this skill lies in its ability to help individuals make sense of their accumulated digital output. Rather than sifting manually through thousands of files, you can use it to intelligently identify and present your own writing, unique insights, and valuable relationship records. This makes it easier to repurpose old ideas, rediscover forgotten projects, or consolidate scattered personal histories. For example, if you've contributed to many open-source projects or written countless internal memos, the skill can help you quickly gather your specific contributions.
It's important to reiterate the boundaries within which this tool operates. The skill is built to enhance your personal archiving efforts through guided discovery, not to automate complex data management tasks without oversight. Its function is to surface items for your review, capturing your reactions to build a personalized understanding of value. It does not perform automated classification without this review step, ensuring that the final decision on content categorization always rests with you. Furthermore, the tool does not interfere with or bypass existing privacy settings or security measures on your system or cloud services. Its operations are confined to the paths you explicitly authorize, adhering to a strict user-defined scope. This ensures that while it helps manage your personal information, it does so within clearly defined and controlled parameters.
Frequently Asked Questions
Q: How does this skill ensure my privacy and data security?
A: It ensures privacy through a mandatory configuration setting, archive-crawler.scan_paths: in gbrain.yml. It will only scan directories you explicitly list. It also relies on your manual review for content classification, never making automated decisions about sensitive data.
Q: What types of files can it process? A: The skill supports plain text, HTML, Markdown, .mbox email, .docx Word documents, .pst Outlook files, and various compressed archives.
Q: Can it automatically delete or modify files in my archive without my explicit instruction? A: No, the tool does not delete or modify original files without your explicit approval during the review phase. It helps organize ingested content into canonical directories based on your filing rules and reactions, leaving original files untouched unless you specifically instruct otherwise through interaction.
This skill offers a structured approach to managing your digital history. It helps you surface valuable information from your personal files on your terms.
