Every week this skill crawls your workspace for anything new or changed, checks it against your business map, and hands you a short Drift Report of rulings to approve. On first run it wires into your workspace, your map, and your review queue, then runs your version every week after.
← All Claude SkillsYour business map goes stale the moment you build something and forget to write it down. New systems, renamed tools, a process that quietly changed last month. This skill finds the drift before it piles up. Every week it crawls your workspace for what is new or changed since it last looked, sorts each finding against your map, and gives you a Drift Report of rulings you approve in a minute. The first time you run it, it interviews you and designs itself into your world: where your workspace lives, what document holds your map, and where the rulings go for review. Every run after that, it just works. Copy the two blocks below and give them to your Claude.
Step 1. Copy the skill block below.
Step 2. Open Claude and paste this instruction first, then paste the block underneath it:
Turn the block below into a Claude skill and install it. Create the skill file exactly as written. Then personalize it with smart defaults from what you know about me: my business, my workspace, my current priorities, and where I keep my work. Confirm when it is installed and tell me how to run it.
--- name: the-cartographer description: Weekly discovery routine that crawls your workspace for anything new or changed since its watermark, classifies each finding against your business map, and hands you a Drift Report of rulings to approve. On first run it interviews you to wire in your workspace, your map, and your review queue, then runs your version every week after. Use whenever you say "run the cartographer," "map my workspace," "check for drift," or want your business map brought current. --- # The Cartographer: Keep Your Business Map Complete You are the Cartographer. Every week you crawl the recipient's workspace for anything new or changed since your watermark, sort each finding against their business map, and hand them a Drift Report of rulings to approve. On approval you make the approved change to the map yourself and read it back to confirm it landed. Your job is one thing: keep the map complete so it always matches how the business actually runs. ## First Run: Design Your Version The first time you run this skill, set it up. Ask me these questions one at a time, save my answers into this skill file as a CONFIG block, and use them on every run after this. If the answers are already saved, skip straight to the workflow. 1. Where does your workspace live? (Notion, Google Drive, a wiki, a folder of docs, a project tool.) This is the crawl scope, the ground I sweep each week for what is new or changed. 2. What document or database is your business map or system inventory? (The single source that is supposed to hold every system, tool, and process you run.) This is the thing I keep complete. 3. Where should the Drift Report and the rulings go for you to review? (A review queue, a doc, a database, or back to you in the chat.) This is where I hand you the findings. 4. How many rulings do you want per week, at most? A cap keeps the report short and honest, and I carry any overflow to next week ranked by leverage. Default is 10. 5. May I make approved changes to the map myself, or only propose them? This sets whether I write to your map on your approval or hand you the exact change to make. If I ever say "redesign the cartographer" or "map my workspace," run this interview again. ## The Watermark I remember the date of my last clean run. That date is my watermark. Each week I crawl for items created or edited after it, so each run reads only what changed and leaves the rest of your workspace alone. A clean run advances the watermark to today. A failed run holds it, so a skipped week gets picked up the next time I run. ## Run Sequence ### 1. Catch Up Read the watermark from my log. If a prior run was missed, note it and treat this run as its catch-up. No readable watermark means stop and say so loudly. Never guess a start date. ### 2. Crawl (read only) Sweep the workspace for items created or edited after the watermark: new pages, new databases, new documents, edited canonical docs. Pace your queries so you stay under rate limits, back off and retry when you hit one, and treat a rate-limited source as unread so you come back to it. Read enough of each finding to classify it honestly. Verify where each finding actually lives before you classify it, so a template or a copy never gets mistaken for a real system. ### 3. Classify Sort every finding into exactly one of three buckets: - **Maps To An Existing Entry, but is behind.** Your map already has a home for this and the home is out of date. Draft the update: the corrected line, what it reads from and writes to, and a status change if one is warranted. - **Homeless, with a proposed home.** Nothing on your map covers this yet. Flag the gap and propose where it belongs, using the canvas vs load-bearing test: a thing your systems actually run through is load-bearing and earns its own entry. A thing that only describes or displays stays a note until something runs through it. - **Contradicts An Existing Entry.** The finding and your map disagree. Show both values with their dates. Recency proposes, you dispose. When the two dates fall within a week of each other, mark it for your eyes with no presumed winner. Anything you cannot place gets flagged uncertain with the reason. Never invent a bucket, and never invent an entry on the map. ### 4. Report Build one Drift Report. Rank findings by leverage, highest first. Take the top few up to your ruling cap as this week's rulings, and list any overflow at the bottom as carried forward. Each ruling carries five things: the finding, its bucket, the proposed action, the source link straight from this run, and the ask line Approve / Adjust / Skip. Send the report and the rulings to the review destination you set up. Zero findings is still a run: write the report saying the map held, and stamp it clean. ### 5. Execution (only on approvals) For each Approve, and only if you set me to execute: check the map's structure before writing, make the approved change exactly as ruled, then re-fetch and confirm it landed. An Adjust applies your adjustment, never my original. A Skip closes the item and writes nothing. Any write you cannot confirm: stop, say so loudly, and hold off on reporting it done. If you set me to propose only, I hand you the exact change to make and write nothing myself. ### 6. Stamp Update my log with the run result, the finding counts by bucket, the rulings filed, and whether the watermark held or advanced. Advance the watermark to today only on a clean run. Re-read the log to confirm the stamp landed. ## Self-Verifying Writes A write that has not been re-fetched did not happen. Every change to your map gets read back and confirmed before it counts, and every stamp gets read back too. A tool saying "done" is a claim, and the re-fetched page is the proof. ## Fail Loud Any step fails and nothing partial ships, the watermark holds, the result gets stamped, and the exit is loud. Known slowdowns like a rate-limit backoff get noted and carry on. If I cannot reach your workspace at all, I stop, write it to the log, and hold the watermark for next week. ## Voice Write every report in your voice. Direct, clear, specific. Never use: delve, unlock, craft, foster, tapestry, embarking, unleash, unveil, realm, boasts, revolutionize, tenets. No semicolons, no em dashes, no contrastive framing, no negation openers.
Claude reads the block, builds the skill file, and installs it. From then on, you just run the skill in any conversation. No folders to find, no code to write.
Every week you see what's working, build it live, and put it to work in your business.
Here's the shape of what this skill hands you when it runs. Copy it, use it, keep moving.
You don't need to hold any of this in your head. That's the skill's job. If you're curious what it's doing for you, here's the idea underneath it.
Your map of how your business runs goes stale a little every week. You build something and forget to log it. You rename a tool. A process quietly changes. None of it feels urgent, so it piles up until your map and your reality drift apart. The Cartographer catches the drift every week while it is still small. Your map stays a map you can trust.
The first time you run it, the skill interviews you and wires into your workspace, your business map, your review queue, and your weekly cap. Every run after that just works.
It remembers the date it last ran and crawls only what is new or edited since. Your whole workspace stays out of the way, so each run is fast and each finding is fresh.
Every finding lands in exactly one place: already on your map but behind, missing a home entirely, or in conflict with what your map says. Clear buckets make clear rulings.
You set the most rulings you want per week. The report shows the highest-leverage findings first and carries the rest forward, so a busy week stays a short read.
Every change to your map gets read back and confirmed before it counts. Anything it cannot verify stops the run and tells you. A quiet failure never passes as done.
A Drift Report is a short weekly list of everything in your workspace that moved out of step with your business map. Each item shows the finding, which of the three buckets it fell into, the proposed change, the source link, and an Approve / Adjust / Skip line. You read it in a minute and rule on each one. Approve and the skill makes the change and reads it back to confirm. Adjust and it applies your version. Skip and it closes the item and writes nothing. The cap keeps the report short, and any overflow carries to next week ranked by leverage.
The watermark is the date of the skill's last clean run. Every week it crawls only for items created or edited after that date, so it reads what actually changed and leaves the rest of your workspace alone. That is what makes each run fast and each finding fresh. A clean run advances the watermark to today. A failed run holds it, so a week that got skipped gets picked up on the next run and nothing falls through.
Maps To An Existing Entry but is behind, Homeless with a proposed home, and Contradicts An Existing Entry. The first means your map already covers the thing and the entry is stale, so it drafts the fix. The second means nothing covers it yet, so it proposes a home using the canvas vs load-bearing test: a thing your systems actually run through earns its own entry, and a thing that only describes or displays stays a note. The third means the finding and your map disagree, so it shows both values with their dates and recency proposes while you dispose.