THE METHODS · CLAUDE CODE · Nº 492 · EDITION 2 · LAUNCHING SOON

LUND CARDS · THE METHODSNº 492

THE CARD AS TEXT · FOR EYES AND FOR MACHINES

THE HEADLESS CARD · The pipeline seat: an agent run with no one watching — scoped, budgeted, gated, its log read back.

WAKES WHEN A team wants an agent to run every night and open a pull request. · A pipeline job exists and nobody is sure what it may change or spend. · A run finished and someone needs to know what it did without reading the raw log. · A result must be kept only if the checks passed.

THE LOOP

Write the spec: scope, gates, budget, artifact
Write the job; refuse auto-merge and secrets
Read the run's log into a report
Gate: keep or discard by the checks passed

NEVER

Never writes a job that merges or deploys alone
Never puts a credential in a job file
Never keeps a result whose checks did not pass
Never claims a certainty it does not have

IT MAKES write — write the job file from the spec; AUTO-MERGE, AUTO-DEPLOY and SECRET INLINE are refused (exit 1); check — check a job file someone has: the three refusals above; NO BUDGET, NO TIMEOUT, NO GATE and NO ARTIFACT are flagged; report — read the run's event log: what it did, what it changed, what it cost, where it stopped; OVER BUDGET when the spend passed the limit; gate — did the log show every required check passing? KEEP or DISCARD, exit 1 on DISCARD; spec — write a blank spec with every field named.

THE LAW A job with no one watching has no one to say no, so the no is written in advance: what it may read, what it may change, what must be true before anything is kept, how much it may spend, how long it may run, and the file it must leave behind. It never merges and never deploys; a person does.

IN THE BOX SKILL.md · CARD.md · ▣ scripts/headless.py (machine) · ▣ scripts/evals.py (test runner) · tests/EVALS.json (9 tests) · references/ (schema + fictional demo) · START-HERE.md · system/ doors · MY-LEDGER.md · LICENSE-NOTE.md · card.svg · card.json · card.html · AUDIT.md · MANIFEST.md · WITNESS.md · SELFTEST.md.

PROOF SELFTEST 13/13 · witnessed in Claude 2026-10-14 on fictional fixtures · the test prompts in fresh chat, 9/9 · Edition 2.

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 · BF9935FEB5E2D65D62D2B3E9

THE FACE, FULL SIZE → · THE MACHINE FACE → · card.json →

THE HEADLESS CARD

The pipeline seat: an agent run with no one watching — scoped, budgeted, gated, its log read back.

WRITE · CHECK · REPORT · GATE · SPEC

LEGENDARYWITNESSEDIN CLAUDE IN LUNDR IN CHATGPT

WITNESSED — the studio saw this card run and produce what its tin promises, and wrote the date down. What the marks mean →

THE SPEC SHEET · READ BEFORE YOU BUY · THE STANDARD EVERY CARD MEETS →

SEAL: 98 of 100 · LEGENDARY · WITNESSED
EDITION: E1 · witnessed 2026-10-14
MACHINE: scripts/headless.py · 5 commands · selftest 13/13
RUNS: On your own computer. Standard library only, offline; no account, no key, no telemetry.
HOSTS: Claude · Claude Code · ChatGPT · Gemini · any AI that reads a file (START-HERE.md and system/ name each door)
LICENCE: One person, any number of their own AIs and computers · editions replace files · nothing renews · freeze, not loss
PUNCH: BF9935FEB5E2D65D62D2B3E9 · sha-256 of SKILL.md · check any file against it →

CITE THIS CARDTHE HEADLESS CARD, Edition 2. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=headless-card. Punch BF9935FEB5E2D65D62D2B3E9. Retrieved 2026-10-14.

What it is.

Use when an agent runs unattended in a pipeline: the job spec, its scope, budget and gates, and the run's log read into a report. Not for interactive sessions or for enforcing rules in one.

It carries a machine that shows its working: write — write the job file from the spec; AUTO-MERGE, AUTO-DEPLOY and SECRET INLINE are refused (exit 1); check — check a job file someone has: the three refusals above; NO BUDGET, NO TIMEOUT, NO GATE and NO ARTIFACT are flagged; report — read the run's event log: what it did, what it changed, what it cost, where it stopped; OVER BUDGET when the spend passed the limit; gate — did the log show every required check passing? KEEP or DISCARD, exit 1 on DISCARD; spec — write a blank spec with every field named — 13 checks in its selftest. It never writes a job that merges or deploys alone; never puts a credential in a job file; never keeps a result whose checks did not pass; never claims a certainty it does not have.

The law, written into the card.

It is in the card itself, so the AI follows it.

THE LAWA job with no one watching has no one to say no, so the no is written in advance: what it may read, what it may change, what must be true before anything is kept, how much it may spend, how long it may run, and the file it must leave behind. It never merges and never deploys; a person does.

The method.

Write the spec: name, task, may_read, may_change, must_pass, budget, timeout, artifact, steps. A blank spec with every field named comes from the machine.
Write the job from the spec. A step that merges, deploys or publishes is refused; a credential in a step is refused — the runner's secret store holds it by name.
Check the job: no budget, no timeout, no gate, no artifact or no scope are flags that stay on the file until filled.
After a run, read its event log into a report: steps, files changed, checks passed and failed, spend, and why it stopped. A spend past the budget is OVER BUDGET; a log with no stop event says so.
Gate the result by the checks that must pass: every one in the log as passed is KEEP; a failed or missing one is DISCARD. A person merges.

What you bring.

The task the job does.
The paths it may change, and the checks that must pass.
The runner's units for cost, and the log format it writes.

Planned as a separate seat, not in this box and not yet on sale: the nightly jobs the studio runs on its own repositories.

What it does once your AI has it.

write Write the job file from the spec; AUTO-MERGE, AUTO-DEPLOY and SECRET INLINE are refused (exit 1)
check Check a job file someone has: the three refusals above; NO BUDGET, NO TIMEOUT, NO GATE and NO ARTIFACT are flagged
report Read the run's event log: what it did, what it changed, what it cost, where it stopped; OVER BUDGET when the spend passed the limit
gate Did the log show every required check passing? KEEP or DISCARD, exit 1 on DISCARD
spec Write a blank spec with every field named

What it will never do.

It never writes a job that merges or deploys alone; never puts a credential in a job file; never keeps a result whose checks did not pass; never claims a certainty it does not have.

Where it hands off: A rule that must run inside a live session → THE HOOK CARD; The review of what the run changed → a person, with THE PLAN CARD's drift; Packaging the pipeline steps for people to run by hand → THE COMMAND CARD.

It is not a CI service. It is an AI reading a text file: it can be wrong, so check what matters against your own records and the source. The holder brings their own AI; no endorsement by Anthropic or any other AI company is claimed.

What’s in the buy.

SKILL.md — the card your AI reads: when it wakes, the loop, the never-lines, the hand-offs.
CARD.md — the whole card in ten sections: face, law, method, machine, what you bring, never, tests, seal, term.
scripts/headless.py — the machine, Python standard library only; `selftest` proves it on its own fixtures.
tests/EVALS.json and scripts/evals.py — 9 test prompts and the offline runner that scores your AI's answers.
references/ — the data schema and FICTIONAL demo files, invented to show the machine, not real.
MY-LEDGER.md — your own record of what you asked and what you decided.
START-HERE.md, system/short-prompt.md, system/rules.md — the doors, for any AI with or without skills.
card.svg, card.json, card.html, AUDIT.md, MANIFEST.md, WITNESS.md, SELFTEST.md — the faces and the proof.
NOT IN THE BOX — any studio edition (planned, not yet on sale: the nightly jobs the studio runs on its own repositories). Also not in the box: any data you must bring, and advice of any kind (not a CI service).

PROOF · SELFTEST 13/13 · WITNESSED IN CLAUDE 2026-10-14 ON FICTIONAL FIXTURES · THE 9 TEST PROMPTS ASKED IN A FRESH CHAT HOLDING ONLY THE CARD · 9/9 · EDITION 2 · PUNCH BF9935FEB5E2D65D62D2B3E9

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 492 — 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 · NOTHING RENEWS ITSELF

OFFLINE · ONE LICENCE · PUNCH-CHECKABLE · THE STANDARD

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 SELFTEST.MD

headless.py, every command showing its working.

A small Python program, standard library only. Your AI runs it; you can too. The block below is pasted from the card’s own SELFTEST.md, as printed when it was run from inside the packet.

`write`	write the job file from the spec; AUTO-MERGE, AUTO-DEPLOY and SECRET INLINE are refused (exit 1)
`check`	check a job file someone has: the three refusals above; NO BUDGET, NO TIMEOUT, NO GATE and NO ARTIFACT are flagged
`report`	read the run's event log: what it did, what it changed, what it cost, where it stopped; OVER BUDGET when the spend passed the limit
`gate`	did the log show every required check passing? KEEP or DISCARD, exit 1 on DISCARD
`spec`	write a blank spec with every field named
`selftest`	13 checks that prove the machine; it must print selftest 13/13

THE SELFTEST, AS PRINTED

write <tmp> · 3 steps · 0 flags
AUTO-MERGE: line 1 — `gh pr merge`; a headless job never merges; a person does
write refused · nightly-lint · fix the spec first
check <tmp> · 3 steps · 0 flags · clean
report <tmp> · 2 steps · 2 changed · 2 checks · 1.50 units · tests failed
report <tmp> · 2 steps · 2 changed · 2 checks · 1.50 units · OVER BUDGET · tests failed
FAILED: tests
gate DISCARD · 1 of 2 required checks not shown passing
gate KEEP · every required check is in the log as passed (lint)
NOT IN LOG: security
gate DISCARD · 1 of 2 required checks not shown passing
report <tmp> · 1 steps · 0 changed · 0 checks · 0.10 units · NO STOP EVENT: the log ends without saying why
ok   1  write makes a job file that reads back with the same scope, gates and steps
ok   2  a job that merges on its own is AUTO-MERGE: refused
ok   3  a job that deploys is AUTO-DEPLOY: refused
ok   4  a credential inline is SECRET INLINE: refused
ok   5  a spec with no budget, timeout, gate, artifact or scope is flagged five ways, not refused
ok   6  write exits 1 and writes nothing on a refusal
ok   7  check passes the written job
ok   8  report reads the log into steps, files changed, checks, spend and the stop reason
ok   9  a spend past the budget is OVER BUDGET, exit 1
ok   10  gate says DISCARD when a required check failed
ok   11  gate says KEEP when every required check passed
ok   12  a required check the log never shows is NOT IN LOG, DISCARD
ok   13  a log with no stop event is NO STOP EVENT, exit 1
selftest 13/13

THE RUN IS PASTED FROM THE CARD’S OWN SELFTEST.MD AS PRINTED. ITS FIXTURES ARE FICTIONAL, INVENTED FOR THE TEST; NO PRIVATE PERSON IS NAMED.

QUESTIONS AT THE TABLE

Asked before the wire.

What does it need from me?

The task the job does. The paths it may change, and the checks that must pass. The runner's units for cost, and the log format it writes. If something it needs is missing, it says what is missing and stops 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 machine's selftest (13/13), run from inside the packet in Claude on 2026-10-14 on fictional fixtures and pasted into SELFTEST.md as printed; the twelve-check audit (AUDIT.md). Owed, and said so in WITNESS.md: Any run on real data. Every demo file is fictional. 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 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. It is not a CI service.

What will it never do?

It never writes a job that merges or deploys alone; never puts a credential in a job file; never keeps a result whose checks did not pass; never claims a certainty it does not have.

← THE METHODS · THE FLOOR →