THE LOAD CARD · Weighs what a skill costs in context: always-on, when triggered, and on demand.
WAKES WHEN Someone asks why a skill is heavy · Someone asks how many skills they can install · A skill needs slimming
THE LOOP
Take the skill or library.
Run `weigh`.
Read FAILs, then WARNs.
Show the always-on total and the heaviest items.
Propose the cut that saves most.
NEVER
Promises exact token counts
Rewrites without asking
Calls a limit current without the edition date
IT MAKES weigh — weigh skills and write LOAD.md and load.json.
THE STANDARD every token figure is labelled an estimate; the always-on cost is reported first for a library; published limits are quoted as published; the machine reports; verify by execution; say the limit out loud in the work; nothing hidden; the person decides.
PROOF SELFTEST 16/16 · witnessed 2026-10-13 on material and live on Claude · Edition 1.
THE SEAL 96 of 100 · LEGENDARY · WITNESSED · Trigger 4 · Machinery 5 · Law 5 · Portability 5 · Proof 5
IT CAN BE WRONG an AI reading a text file: it can be wrong and can invent a number, a rule, a name or a date; check what matters against your own records and the source; the risk is yours. LICENSE-NOTE.md rides in the packet and is the licence.
THE PUNCH · SHA-256 OF SKILL.MD · 9F92E882118EA65082F612AF
CITE THIS CARDTHE LOAD CARD, Edition 1. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=load-card. Punch 9F92E882118EA65082F612AF. Retrieved 2026-10-13.
What it is.
Weighs what a skill or a whole skill library costs in an AI's context, tier by tier: the name and description every session pays for, the SKILL.md body paid when it triggers, and files read on demand; flags long bodies, orphan references, deep reference chains and broken links.
Use when someone asks why a skill is heavy, how many skills they can install, or how to slim one down. Do not use to count tokens in a chat or a document (that is a different job) or to check security.
It carries a machine that shows its working: weigh — weigh skills and write LOAD.md and load.json — 16 checks in its selftest. It will never promises exact token counts; never rewrites without asking; never calls a limit current without the edition date.
Eight laws, written into the card.
They are in the card itself, so the AI follows them.
EVERY TOKEN FIGURE IS LABELLED AN ESTIMATEEvery token figure is labelled an estimate.
THE ALWAYS-ON COST IS REPORTED FIRST FOR A LIBRARYThe always-on cost is reported first for a library.
PUBLISHED LIMITS ARE QUOTED AS PUBLISHEDPublished limits are quoted as published, with the date of the edition.
THE MACHINE REPORTSThe machine reports; the person cuts.
VERIFY BY EXECUTIONVerify by execution, not by reasoning: run the machine and read what it prints before saying a thing is done.
SAY THE LIMIT OUT LOUD IN THE WORKSay the limit out loud in the work: every result names what it did not check.
NOTHING HIDDENNothing hidden: two readers, the person and the AI, see the same words.
THE PERSON DECIDESThe person decides.
What rides in the card.
In the card’s own voice, with a machine that shows its working.
THE GUIDEreferences/THE-TIERS.md
THE MACHINEload.py · one commandweigh — weigh skills and write LOAD.md and load.json. Every command prints its working and refuses a bad input; 16 checks in its selftest.
What it does once your AI has it.
weigh Weigh skills and write LOAD.md and load.json
On the record this cut's own 18 packets — real material (not made up for the test). First run: 109 warnings, almost all 'a reference links to another reference' — the packets' CARD.md, TIN.md and START-HERE.md pointed at SELFTEST.md, WITNESS.md and MANIFEST.md, which SKILL.md did not name. A real finding about the cutter, not the cards' words. - After the fixes below: 0 FAIL, 0 WARN across all 18. Always-on cost of installing all 18 is printed in LOAD.md. Live on Claude Sonnet, 2026-10-13: 7 of 7 passed; the card fired on 2 of 3 trigger cases and on none of its hand-off cases.
What it will never do.
It will never promises exact token counts; never rewrites without asking; never calls a limit current without the edition date.
Estimates, not a tokenizer. It cannot see how a particular model or host packs context, and it does not judge whether the words are good. It is an AI reading a text file: it can be wrong, so check what matters against your own records and the source.
What’s in the buy.
CARD.md — the whole card: its boot, laws, loop, when it says no, its seals, ledger, term and score · SKILL.md — the door your AI reads first · START-HERE.md — the first five minutes, for you
THE-TIERS.md — the guide in references/, in the card's own voice
scripts/load.py — the machine: weigh, standard library only, no network, selftest 16/16
WITNESS.md — the runs on real material and live on Claude, with what was not run · SELFTEST.md — the machine's own test, as run at the cut · MANIFEST.md, every file hashed · CHANGELOG.md
tests/evals.json + EVALS.md — the cases it must take, hand off and refuse
MY-LEDGER.md — the card's memory, on your own machine · TIN.md, the nine lines on the tin · LICENSE-NOTE.md
card.svg + card-machine.svg + card.json · card.html, the card as one offline file · doors/ — the rules files for Cursor, Codex, Gemini, Copilot and Claude Code, the system text for any other AI, a local-model file and an MCP door
PROOF · SELFTEST 16/16 · WITNESSED ON MATERIAL · LIVE ON CLAUDE SONNET 2026-10-13 · 7 OF 7 PASSED · EDITION 1 · PUNCH 9F92E882118EA65082F612AF
ON EVERY DOOR · A CARD FOR YOUR AI, A FILE FOR YOUR COMPUTER
In Claude: turn on code execution under Settings › Capabilities, then Customize › Skills › + › Create skill, and upload the zip as it is — START-HERE.md in the box walks you through it, and tells you how to keep the ledger. In ChatGPT, Gemini and the rest: a Project, a GPT or a Gem with the card’s files attached and the short door in the box pasted as its instructions. The words run anywhere an AI reads Markdown; the machines run wherever the host runs Python. Witnessed in Claude only. The host marks →
Take it.
Nº 1194 PROPOSED — LAUNCHING SOON
Sold as is under the term printed. A card instructs an AI you operate; results depend on your model, your files and your judgement. Not medical, legal, tax or financial advice, never the person’s boss, and not a licence to practise. The terms →
One payment. The file is yours to keep and keeps working at its edition; new editions issued in the year after you buy reach you by email as replacement files. Tax is added at checkout where it applies. Nothing auto-renews.
A TERM, NOT A SUBSCRIPTION: NOTHING RENEWS ITSELF AND NOBODY IS CHARGED AGAIN. WHEN A TERM ENDS THE FILES STAY ON YOUR COMPUTER AND STAY READABLE — FREEZE, NOT LOSS.
THE MACHINE · IN THE OPEN · FROM THE CARD’S OWN WITNESS.MD
load.py, every command showing its working.
A small Python program, standard library only. Your AI runs it; you can too. The blocks below are pasted from the card’s own WITNESS.md, as printed.
weigh
weigh skills and write LOAD.md and load.json
selftest
16 checks that prove the machine; it must print 16/16
THE COMMANDS, IN ORDER, ON THIS CUT'S OWN 18 PACKETS
THE RUNS ARE PASTED FROM THE CARD’S OWN WITNESS.MD AS PRINTED. FIXTURES MARKED FICTIONAL ARE THE STUDIO’S OWN EXAMPLES, INVENTED FOR THE RECORD; NO PRIVATE PERSON IS NAMED.
QUESTIONS AT THE TABLE
Asked before the wire.
What does it need from me?
The skill folders or zips. Take the skill or library, and it goes from there.
Does it send anything anywhere?
No. The machine is a standard-library Python program that runs on your own computer and sends nothing anywhere; the words run wherever an AI reads Markdown. No server, no account, no key.
What is on the record?
The self-test (16/16), a run on real or invented material with every command and verdict pasted into WITNESS.md as printed, and live runs on Claude Sonnet: 7 of 7 passed; the card fired on 2 of 3 trigger cases and on none of its hand-off cases.; the punch on the face is the SHA-256 of SKILL.md, so a downloaded copy can be checked against this room.
Can it be wrong?
Yes — it is an AI reading a text file. Estimates, not a tokenizer. It cannot see how a particular model or host packs context, and it does not judge whether the words are good.
What will it never do?
It will never promises exact token counts; never rewrites without asking; never calls a limit current without the edition date.
