I took 4,305 screenshots last year.

I am an absolute screenshot fiend.

I captured 4,305 in 2025 alone.

I intend to treat them as my ‘lazy bookmarks’. Something cool, useful or interesting that I should refer back to later.

But I never do. Nothing happens with them. They pile up on my phone never to be seen again.

I do the same thing with bookmarks in my browser. And starred emails in my inbox. Pinterest boards. Even my camera roll is full of photos of interesting things I see out on the street.

I’ve always dreamed of being able to organise these resources properly. I know for a fact that I’m sitting on a wealth of useful knowledge & inspiration that I’m doing nothing with.

Years of my own curated ‘taste’ - with no easy way to access it.

There are a few apps that claim to take care of this for you. But they’re super labour intensive. They’re basically glorified cloud storage - where you become the librarian. You’re the one responsible for tagging, filing and cross referencing everything manually. The whole system falls apart the second you get lazy. And let’s face it - you’re going to get lazy.

This is not a time problem. It is an invisible asset problem. The references are already there. The cost is that the asset is dormant - something you already own that you cannot harness.

That is what this week’s skill goes after.

GOING IN SEARCH OF A LIBRARIAN

First we need a single location where all your disorganised assets can land. Could be a Dropbox folder, a Google Drive - it doesn’t really matter. Just a central location in the cloud you can access across all your devices so that all your assets - regardless of their source - can end up in one place.

This folder becomes the ‘goods receiving’ for our archive. Now we need a dedicated ‘librarian’ who is going to trawl through each and every asset in meticulous detail, understand the content each one, tag it, and file in such a way that we can easily search for it later.

Then - when you’re looking for some knowledge recommendations. Simply chat to your librarian and (thanks to their encyclopaedic knowledge of our entire archive) they will be able to dig out exactly the references you’re looking for.

And to take things even further. Over time, your librarian will start to notice patterns, trends and create links between assets. Turning your archive into this living, breathing fountain of knowledge.

And yes, you guessed it - this is the kind of thing AI excels at.

WHAT I BUILT

An agentic skill called The Agent Inbox.

You setup a folder in your cloud storage of choice, and point your agent at it. It runs best inside Claude Cowork or Claude Code - but is also MCP compatible.

When you drop a file in (an image, a screenshot, a video, a voice note, a PDF, a saved article URL, plain text) and run the ‘process inbox’ command, the skill picks each file up, reads it, understands it and files it in a Notion database.

(I built this to work for Notion - but it could quite easily be adapted to work with tools like Airtable if you prefer)

The skill extracts all the information it can reasonably interpret. Colour, mood, author, topic, theme, the list goes on. All this information is logged alongside each asset to help with later discover via search.

The Notion database essentially becomes your own mini search engine for your files.

The skill is also able to improve the filing system over time. Alongside the Notion database, it also keeps its own record of the internal filing taxonomy you have ended up with. Patterns it finds. Feedback you’ve given.

Any time the skill isn’t sure about something - or needs a bit more context on an asset - it will ask you. Over time, it learns your personal preferences, draws its own connections and will need to ask you less and less.

PRO TIP: You can provide your agent with some more context on an asset by renaming the file itself. Eg: 'I like the styling of this website, something to look at when I launch my new blog.png’

SETUP

(one-time, takes about ten minutes)

1. Pick your cloud source

The skill ships with adapters for Dropbox, Google Drive, and Mega. Point the skill at one folder inside whichever you choose. That is your inbox.

2. Connect your Notion database

The skill creates the database for you on first run. Multi-select tag columns for PARA, Type, and free-form Tags. Standard fields for title, summary, source, capture date, and a thumbnail or preview block. It’s deliberately built as a single flat database to make it easier to search. But if you’re a Notion power user, you can easily build your own views and filters on top of it (although this isn’t required).

3. Set your filing rules

By default, the skill will organise the files in your inbox to a parallel ‘processed’ folder once they have been added to the database. That means your Inbox folder won’t get cluttered with thousands of files as you add to it over time.

Alternatively, you can have the skill delete a file once it has extracted the information from it (eg screenshots, website links etc) - but this isn’t recommended.

IN USE

(run whenever you have collected things in the inbox, or on a schedule)

1. Drop files in the folder

Through any normal route. Desktop or mobile. Save a screenshot to a synced folder. Drag a PDF in. Drop a voice memo. Forward an article URL. Anything goes.

2. Run ‘process inbox’

Run the command. The skill checks the folder and processes any new files.

For images and screenshots, Claude vision reads the contents and proposes tags. For voice notes and short videos, Claude transcribes the audio and extracts the substance. For PDFs and articles, it reads and summarises. For anything containing a URL - it browses that web page to extract all the relevant information and context.

For each file, you see the proposed tags before the row gets written. Accept, edit, or reject. Anything you reject or give feedback on speeds up the self-learning.

3. Search and pull

Once the inbox has been running for a while, the Notion database becomes a real asset. Filter by PARA. Filter by tag. Search by colour or mood or topic. When you sit down to write a brief, draft a deck, plan a shoot, build a moodboard - you have a simple but powerful way to tap into your archive.

PRO TIP: Put it on a schedule. You can setup a Claude Routine (or cron job) to have the skill check the inbox periodically so you can make this totally hands off (as long as your computer is running). I have mine run at midnight each day.

GREAT FOR TEAMS

This doesn’t only apply to your personal archive. The infrastructure scales for small teams too.

Point a shared folder at a shared Notion database. Every team member drops what catches their eye in. The shared database becomes the team’s living reference vault, searchable by anyone, no Slack archaeology, no “where did Alex post that thing in March”.

“WAIT, WON’T THIS DULL MY EYE?”

Worth addressing. The fear is that letting an AI system tag your archive thins out the deep familiarity that comes from manually re-encountering a saved thing. You stop being the curator and start being the operator.

The honest answer: it is the inverse. Tagging the archive surfaces what you have actually saved, by category, at scale. It allows you to build up rich patterns and connections that you would be unlikely to spot before.

The thing you lose is the friction. The thing you gain is the legibility of your own taste.

WHERE TO START

I’ve supplied the full skill below. Setup is roughly ten minutes - point it at your cloud folder, connect your Notion, set your rules. After that it just runs.

If you would prefer to build this yourself, the playbook is above. Pick a cloud source. Build a Notion DB. Write the watcher. Write the vault-context loop. It will take you about an afternoon and a few weeks of calibration. The skill above just saves you that time, and gives you the version I have already calibrated against my own archive.

Catch you next week!

Sam

P.S. If you are a business owner or work in a small creative team who is looking to expand their AI capabilities, I’m opening up 3 x free slots this week for my new ‘Creative AI Audit’ offer. If you’d like to be one of my guinea pigs - shoot me an email at hello@madewithmachines.com or reply to this email.