THE CLASH CARD · Finds the skills in a library that compete for the same requests, and whether they hand off.
WAKES WHEN A skill library grows past a handful · Two skills seem to fight over requests · Before a new skill is added
THE LOOP
Take the library.
Run `check` and list CLASH pairs first.
Run `route` on sample requests.
Draft the hand-off sentence for each clash.
Re-run until no CLASH is left.
NEVER
Calls an estimate the model's choice
Renames skills on its own
Hides a clash under a lower threshold
IT MAKES check — compare every pair of descriptions; write CLASH.md and; route — estimate which skill each prompt would pick; write.
THE STANDARD overlap is measured; a word-overlap estimate is never called…; every CLASH gets a drafted hand-off sentence; nothing is renamed or deleted without…; verify by execution; say the limit out loud in the work; nothing hidden; the person decides.
PROOF SELFTEST 15/15 · witnessed 2026-10-13 on material and live on Claude · Edition 1.
THE SEAL 98 of 100 · LEGENDARY · WITNESSED · Trigger 5 · 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 · 4F104EF0CB22A9DC60C8E5D2
CITE THIS CARDTHE CLASH CARD, Edition 1. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=clash-card. Punch 4F104EF0CB22A9DC60C8E5D2. Retrieved 2026-10-13.
What it is.
Finds skills in a library whose descriptions compete for the same requests: compares every pair, shows the shared words, and says whether each pair hands off to the other or clashes; also flags descriptions with no do-not clause and routes sample prompts to see which ones are ambiguous.
Use when a skill library grows, two skills seem to fight, or before adding a new skill. Do not use to measure the model's real choice (THE TRIGGER TEST CARD runs that) or to rename skills.
It carries a machine that shows its working: check — compare every pair of descriptions; write CLASH.md and; route — estimate which skill each prompt would pick; write — 15 checks in its selftest. It will never calls an estimate the model's choice; never renames skills on its own; never hides a clash under a lower threshold.
Eight laws, written into the card.
They are in the card itself, so the AI follows them.
OVERLAP IS MEASUREDOverlap is measured, not guessed.
A WORD-OVERLAP ESTIMATE IS NEVER CALLED…A word-overlap estimate is never called the model's real choice.
EVERY CLASH GETS A DRAFTED HAND-OFF SENTENCEEvery CLASH gets a drafted hand-off sentence, approved by the person.
NOTHING IS RENAMED OR DELETED WITHOUT…Nothing is renamed or deleted without the person's word.
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/FIXING-A-CLASH.md
THE MACHINEclash.py · two commandscheck — compare every pair of descriptions; write CLASH.md and; route — estimate which skill each prompt would pick; write. Every command prints its working and refuses a bad input; 15 checks in its selftest.
What it does once your AI has it.
check Compare every pair of descriptions; write CLASH.md and
route Estimate which skill each prompt would pick; write
On the record this cut's 18 descriptions beside the public example skills on this host (60 descriptions, 1,770 pairs) — real material (not made up for the test). No CLASH among the 60. One GUARDED pair (the two browser skills), each naming the other in its do-not clause — checked by hand. 24 public example skills carry no do-not clause, flagged WARN; none of the 18 cards is among them. Live on Claude Sonnet, 2026-10-13: 7 of 7 passed; the card fired on 3 of 3 trigger cases and on none of its hand-off cases.
What it will never do.
It will never calls an estimate the model's choice; never renames skills on its own; never hides a clash under a lower threshold.
Word overlap and tf-idf are a proxy. Two descriptions can share no words and still compete in a model's judgement, and the reverse. 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
FIXING-A-CLASH.md — the guide in references/, in the card's own voice
scripts/clash.py — the machine: check · route, standard library only, no network, selftest 15/15
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 15/15 · WITNESSED ON MATERIAL · LIVE ON CLAUDE SONNET 2026-10-13 · 7 OF 7 PASSED · EDITION 1 · PUNCH 4F104EF0CB22A9DC60C8E5D2
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 1195 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
clash.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.
check
compare every pair of descriptions; write CLASH.md and
route
estimate which skill each prompt would pick; write
selftest
15 checks that prove the machine; it must print 15/15
THE COMMANDS, IN ORDER, ON THIS CUT'S 18 DESCRIPTIONS BESIDE THE PUBLIC EXAMPLE SKILLS ON THIS HOST (60 DESCRIPTIONS, 1,770 PAIRS)
$ python3 scripts/clash.py check clash-lib --out out
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 library of skills; optionally a list of real requests. Take the 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 (15/15), 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 3 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. Word overlap and tf-idf are a proxy. Two descriptions can share no words and still compete in a model's judgement, and the reverse.
What will it never do?
It will never calls an estimate the model's choice; never renames skills on its own; never hides a clash under a lower threshold.
