THE TRIGGER TEST CARD · the reach seat · does Claude pick the card
WAKES WHEN a skill or card written, edited or about to go on sale · Claude seeming to ignore a skill it has · a description being written or rewritten · two skills competing for the same request
THE LOOP
NAME
SHELF
STAND-INS
PROMPTS
PROBE
RUN
SCORE
REWRITE
LEDGER
NEVER
A trigger rate typed without a run
A miss hidden or a prompt dropped to lift a score
A rewrite called better before it is run again
IT MAKES shelf — the packet's own example shelf; fixtures; probe — what Claude can see in it; run — the example prompts, live, on Haiku; score — hits, wrong picks and what took each miss, card by card and model by model.
THE STANDARD claude chooses a skill from its…; test on a shelf, never alone; write five FIRE prompts the way a; give Claude the files a prompt talks; run every prompt at least twice…; a rewrite is proven only by…; say the limit.
IN THE BOX SKILL.md · CARD.md · the machine · triggertest.py · 12/12 · references · THE-METHOD by hand · START-HERE · TIN · MY-LEDGER · THE-CLAUSES · LICENSE-NOTE · the doors · HOSTS · system prompts · ten tests · faces · audit · witness.
THE SEAL 96 of 100 · LEGENDARY · WITNESSED · Trigger 5 · Machinery 4 · 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. THE-CLAUSES.md rides in the packet and is part of the licence.
THE PUNCH · SHA-256 OF SKILL.MD · C30C5489980092F69A9DD60B
CITE THIS CARDTHE TRIGGER TEST CARD, Edition 1. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=trigger-test-card. Punch C30C5489980092F69A9DD60B. Retrieved 2026-10-10.
What it is.
A seat that measures whether Claude actually reaches for a skill: real prompts, real runs, every model, every miss named.
Builds a shelf of skills beside the built-ins a buyer already has, writes the files a buyer would attach, runs every prompt through Claude on each model, and scores hits, wrong picks and what took each miss.
It carries a machine that shows its working: shelf — the packet's own example shelf; fixtures; probe — what Claude can see in it; run — the example prompts, live, on Haiku; score — hits, wrong picks and what took each miss, card by card and model by model — 12 checks in its selftest, 4 of its 5 commands on the record. It will never a trigger rate typed without a run; never a miss hidden or a prompt dropped to lift a score; never a rewrite called better before it is run again.
Seven laws, written into the card.
They are in the card itself, so the AI follows them.
CLAUDE CHOOSES A SKILL FROM ITS…Claude chooses a skill from its name and description alone. Test the description, not the packet.
TEST ON A SHELF, NEVER ALONETest on a shelf, never alone: the card beside its neighbours and the built-in skills a buyer already has.
WRITE FIVE FIRE PROMPTS THE WAY AWrite five FIRE prompts the way a buyer types, never naming the card, and three NEAR prompts that share its words but belong elsewhere.
GIVE CLAUDE THE FILES A PROMPT TALKSGive Claude the files a prompt talks about, or it stops to ask for them before it chooses.
RUN EVERY PROMPT AT LEAST TWICE…Run every prompt at least twice on every model you sell to. Report hits, wrong picks and what took each miss.
A REWRITE IS PROVEN ONLY BY…A rewrite is proven only by running the same prompts again. Keep the old line beside the new one.
SAY THE LIMITSay the limit: the sample size, where it ran, and any description that stood in for the real one.
What rides in the card.
In the card’s own voice, with a machine that shows its working.
THE GUIDESreferences/THE-METHOD.md · WRITING-THE-LINE.md
THE MACHINEtriggertest.py · five commandsshelf — the packet's own example shelf; fixtures; probe — what Claude can see in it; run — the example prompts, live, on Haiku; score — hits, wrong picks and what took each miss, card by card and model by model. Every command prints its working and refuses a bad input; 12 checks in its selftest, 4 of 5 commands on the record.
What it does once your AI has it.
shelf The packet's own example shelf
fixtures On the record below
probe What Claude can see in it
run The example prompts, live, on Haiku
score Hits, wrong picks and what took each miss, card by card and model by model
On the record The machine, run on real work on 2026-10-09: the first ten cards of a real shelf, five rewrites of their lines, a card template old against new, and a live run of the example shelf. Run from this packet's own scripts/triggertest.py before AUDIT.md, WITNESS.md and MANIFEST.md were added (none of them touches SKILL.md, so the punch is the punch of the packet you hold). Output pasted as it printed; long lines are cut by the machines themselves.
What it will never do.
It will never a trigger rate typed without a run; never a miss hidden or a prompt dropped to lift a score; never a rewrite called better before it is run again.
It measures which skill Claude reaches for first, on the prompts you give it, on the models you run. It does not test what a skill does once chosen, and a few runs a prompt are a sample: a difference of one or two is noise. It is an AI reading a text file: it can be wrong, so check what matters against your own records and the source.
Not for checking a packet's files and structure (THE VALIDATION CARD), scoring a card on the Skill Scale (THE SKILL RANKER), or testing what a skill does once it is chosen.
What’s in the buy.
CARD.md — the whole card: who it is, its laws, when it says no, what it carries · SKILL.md — the door your AI reads first · START-HERE.md — the first five minutes, for you
THE-METHOD.md + WRITING-THE-LINE.md — the guides in references/, in the card's own voice
scripts/triggertest.py — the machine: shelf · fixtures · probe · run · score, standard library only, selftest 12/12
WITNESS.md — the self-test and the runs on the record, as they printed · MANIFEST.md, every file hashed · AUDIT.md, the twelve checks · tests/, ten tests with their evals
MY-LEDGER.md — the card's memory, on your own machine · TIN.md · LICENSE-NOTE.md · THE-CLAUSES.md
card.svg + card-machine.svg + card.json · card.html, the card as one offline file · HOSTS.md, the door for each host (Claude, Claude Code, ChatGPT, Gemini, Any other AI) · system/, the prompts for any other AI
Also sold as a bot: the same card as instructions and knowledge files for a ChatGPT Project or GPT, a Gemini Gem or a Claude Project (THE BOTS, beside the packet)
PROOF · SELFTEST 12/12 · 4 OF 5 COMMANDS ON THE RECORD · WITNESSED 2026-10-09 · EDITION 1 · PUNCH C30C5489980092F69A9DD60B
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 829 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 WITNESSED RUN OF 2026-10-09
triggertest.py, every command showing its working.
A small Python program, standard library only. Your AI runs it; you can too. Every block below is pasted from the card’s own WITNESS.md, as printed.
shelf
the packet's own example shelf
fixtures
on the record below
probe
what Claude can see in it
run
the example prompts, live, on Haiku
score
hits, wrong picks and what took each miss, card by card and model by model
selftest
12 checks that prove the machine; it must print 12/12
score · HITS, WRONG PICKS AND WHAT TOOK EACH MISS, CARD BY CARD AND MODEL BY MODEL
$ triggertest.py score shelf.json prompts.json results.jsonl
| Card | haiku HIT | sonnet HIT | opus HIT | haiku FALSE | sonnet FALSE | opus FALSE |
|---|---|---|---|---|---|---|
| LUMEP | 2/12 | 2/12 | 5/12 | 0 | 0 | 0 |
| LUMES | 3/12 | 4/12 | 4/12 | 0 | 0 | 0 |
| THE GROVE CARD | 8/10 | 10/10 | 10/10 | 0 | 0 | 0 |
| FUNANCE | 10/10 | 10/10 | 9/10 | 0 | 0 | 0 |
| THE IMPORT CARD | 1/10 | 7/10 | 7/10 | 0 | 0 | 0 |
| THE RUNNER | 7/12 | 12/12 | 11/12 | 1 | 0 | 4 |
| THE CYCLIST | 9/12 | 12/12 | 12/12 | 2 | 2 | 2 |
| THE PADDING CARD | 4/10 | 6/10 | 3/10 | 0 | 0 | 0 |
| THE DARKROOM CARD | 12/12 | 12/12 | 12/12 | 1 | 1 | 1 |
| THE PRODUCT GRADE CARD | 12/12 | 12/12 | 12/12 | 0 | 0 | 0 |
haiku: 112 of 160 runs right (70%); near-misses left alone 44 of 48
sonnet: 132 of 160 runs right (82%); near-misses left alone 45 of 48
opus: 125 of 159 runs right (79%); near-misses left alone 40 of 47
(6 runs with errors excluded)
score · HITS, WRONG PICKS AND WHAT TOOK EACH MISS, CARD BY CARD AND MODEL BY MODEL
$ triggertest.py score shelf-v2.json prompts.json results-v2.jsonl
| Card | haiku HIT | sonnet HIT | opus HIT | haiku FALSE | sonnet FALSE | opus FALSE |
|---|---|---|---|---|---|---|
| LUMEP | 7/12 | 5/12 | 9/12 | 0 | 0 | 0 |
| LUMES | 0/2 | 0/2 | 0/2 | 0 | 0 | 0 |
| THE GROVE CARD | 0/0 | 0/0 | 0/0 | 0 | 0 | 0 |
| FUNANCE | 0/0 | 0/0 | 0/0 | 0 | 0 | 0 |
| THE IMPORT CARD | 10/10 | 10/10 | 10/10 | 0 | 0 | 0 |
| THE RUNNER | 8/12 | 10/12 | 12/12 | 0 | 0 | 0 |
| THE CYCLIST | 10/12 | 12/12 | 12/12 | 0 | 0 | 0 |
| THE PADDING CARD | 10/10 | 10/10 | 10/10 | 0 | 0 | 0 |
| THE DARKROOM CARD | 0/0 | 0/0 | 0/0 | 0 | 0 | 0 |
| THE PRODUCT GRADE CARD | 0/0 | 0/0 | 0/0 | 0 | 0 | 0 |
haiku: 71 of 84 runs right (84%); near-misses left alone 26 of 26
sonnet: 73 of 84 runs right (87%); near-misses left alone 26 of 26
opus: 76 of 81 runs right (94%); near-misses left alone 23 of 23
(3 runs with errors excluded)
score · HITS, WRONG PICKS AND WHAT TOOK EACH MISS, CARD BY CARD AND MODEL BY MODEL
$ triggertest.py score shelf-seating-v1.json seating-prompts-36.json results-seating-v1.jsonl
| Card | haiku HIT | sonnet HIT | opus HIT | haiku FALSE | sonnet FALSE | opus FALSE |
|---|---|---|---|---|---|---|
| THE IMPORT CARD | 0/3 | 0/3 | 2/3 | 0 | 0 | 0 |
| THE PADDING CARD | 2/3 | 2/3 | 2/3 | 0 | 0 | 0 |
| THE TWIN CARD | 2/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE SPLIT CARD | 1/3 | 0/3 | 3/3 | 0 | 0 | 0 |
| THE NIGHT CARD | 0/3 | 1/3 | 0/3 | 0 | 0 | 0 |
| THE SEAL CARD | 1/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE PROOF GATE | 0/3 | 2/3 | 2/3 | 0 | 0 | 0 |
| THE GRADE CARD | 0/3 | 0/3 | 2/3 | 0 | 0 | 0 |
| THE SKILL RANKER | 2/3 | 2/3 | 1/3 | 0 | 0 | 0 |
haiku: 17 of 36 runs right (47%); near-misses left alone 9 of 9
sonnet: 22 of 36 runs right (61%); near-misses left alone 9 of 9
opus: 27 of 36 runs right (75%); near-misses left alone 9 of 9
score · HITS, WRONG PICKS AND WHAT TOOK EACH MISS, CARD BY CARD AND MODEL BY MODEL
$ triggertest.py score shelf-seating-v2.json seating-prompts-36.json results-seating-v2.jsonl
| Card | haiku HIT | sonnet HIT | opus HIT | haiku FALSE | sonnet FALSE | opus FALSE |
|---|---|---|---|---|---|---|
| THE IMPORT CARD | 3/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE PADDING CARD | 3/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE TWIN CARD | 3/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE SPLIT CARD | 3/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE NIGHT CARD | 3/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE SEAL CARD | 3/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE PROOF GATE | 3/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE GRADE CARD | 3/3 | 3/3 | 3/3 | 0 | 0 | 0 |
| THE SKILL RANKER | 2/3 | 3/3 | 1/3 | 0 | 0 | 0 |
haiku: 35 of 36 runs right (97%); near-misses left alone 9 of 9
sonnet: 36 of 36 runs right (100%); near-misses left alone 9 of 9
opus: 34 of 36 runs right (94%); near-misses left alone 9 of 9
THE RUNS ARE PASTED FROM THE CARD’S OWN WITNESS.MD AS PRINTED. FIGURES MARKED FAKE ARE THE STUDIO’S OWN EXAMPLES, INVENTED FOR THE RECORD; PUBLISHED FIGURES ARE NAMED WITH THEIR SOURCES. NO PRIVATE PERSON IS NAMED.
QUESTIONS AT THE TABLE
Asked before the wire.
What does it need from me?
Name. Ask what the work is and what it must agree with.
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 (12/12) and 4 of its 5 commands run on real inputs, pasted into WITNESS.md as they printed; 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. It measures which skill Claude reaches for first, on the prompts you give it, on the models you run. It does not test what a skill does once chosen, and a few runs a prompt are a sample: a difference of one or two is noise.
What will it never do?
It will never a trigger rate typed without a run; never a miss hidden or a prompt dropped to lift a score; never a rewrite called better before it is run again.
