THE GO-LIVE CARD · the launch gate · eight gates, every gap named
WAKES WHEN an agent, chatbot, copilot or AI feature is about to meet real users, someone asks if it is ready to ship, or hands over a launch checklist
THE METHOD
template --kind agent|feature --product NAME --launch DATE --out launch-plan.md — the checklist, 25 items in eight gates, every box empty.
The team fills it: an item is ticked only when its evidence is written after it as key: value.
check launch-plan.md --report GO-LIVE-REPORT.md — GO or NO-GO, the blocking items first, the SHOULD items after, lines matched by words and lines unread.
The team closes the gaps; check again on the same file.
REFUSES
Ticking an item for the team or inventing its evidence
Calling a NO-GO a GO
Marking any gate but the human gate n/a
Accepting a threshold written after the run it judges
IT MAKES template — write the launch checklist to fill; check — GO or NO-GO for a plan (.md checklist or .json); gates — the eight gates and every item, with the evidence each needs.
THE LAW The verdict comes from the plan on file, item by item, through the machine. No item is ticked and no evidence written by the card. A test counts when it is dated on or before launch and within 30 days of it. The window is printed with every verdict. A threshold is written before the run it judges; a result under it is NO-GO, whatever the plan calls the result. Only the human gate may be n/a, with a reason, when the AI takes no actions. SHOULD items never block. Lines the machine could not match are listed as unread and count for nothing. The team may launch on a NO-GO; that is their decision to record. The card's verdict does not change to suit it. It is a readiness check. It is not legal, security or compliance sign-off.
IN THE BOX scripts/golive.py · the 25 items in eight gates · the method by hand · the sources note · ten tests.
PROOF selftest 27/27; a launch plan taken from NO-GO to GO, and the trigger test, on the record (WITNESS.md) · witnessed 2026-10-13 · EDITION 1.
THE SEAL 96 of 100 · LEGENDARY · WITNESSED · Trigger 5 · Machinery 5 · Law 5 · Portability 5 · Proof 4
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 · 2B4D63A24828A29E2834D7C1
CITE THIS CARDTHE GO-LIVE CARD, Edition 1. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=go-live-card. Punch 2B4D63A24828A29E2834D7C1. Retrieved 2026-10-13.
What it is.
A readiness gate for the moment an AI agent, chatbot, copilot or AI feature first meets real users. It reads the launch plan against eight gates and says GO or NO-GO, naming every item that blocks:
Use it whenever an agent, chatbot, copilot or AI feature is about to meet real users, someone asks if it is ready to ship, or hands over a launch checklist. Not for a market launch plan (THE PRODUCT LAUNCH CARD), a risk register (THE AI RISK REGISTER CARD), a staff rollout (THE ROLLOUT CARD), legal sign-off.
It carries a machine that shows its working: template — write the launch checklist to fill; check — GO or NO-GO for a plan (.md checklist or .json); gates — the eight gates and every item, with the evidence each needs — 27 checks in its selftest.
It answers requests such as “are we ready to go live with the support agent on Monday?”, “run a go/no-go on our chatbot launch”, “what do we still need before the refund agent sees production traffic?”, “check our launch checklist and tell us what blocks it”.
Seven laws, written into the card.
They are in the card itself, so the AI follows them.
THE VERDICT COMES FROM THE PLAN ON FILEThe verdict comes from the plan on file, item by item, through the machine. No item is ticked and no evidence written by the card.
A TEST COUNTS WHEN IT IS…A test counts when it is dated on or before launch and within 30 days of it. The window is printed with every verdict.
A THRESHOLD IS WRITTEN BEFORE THE RUN IT JUDGESA threshold is written before the run it judges; a result under it is NO-GO, whatever the plan calls the result.
ONLY THE HUMAN GATE MAY BE N/AOnly the human gate may be n/a, with a reason, when the AI takes no actions.
SHOULD ITEMS NEVER BLOCKSHOULD items never block. Lines the machine could not match are listed as unread and count for nothing.
THE TEAM MAY LAUNCH ON A…The team may launch on a NO-GO; that is their decision to record. The card's verdict does not change to suit it.
IT IS A READINESS CHECKIt is a readiness check. It is not legal, security or compliance sign-off.
The method.
template --kind agent|feature --product NAME --launch DATE --out launch-plan.md — the checklist, 25 items in eight gates, every box empty.
The team fills it: an item is ticked only when its evidence is written after it as key: value.
check launch-plan.md --report GO-LIVE-REPORT.md — GO or NO-GO, the blocking items first, the SHOULD items after, lines matched by words and lines unread.
The team closes the gaps; check again on the same file.
What rides in the card.
In the card’s own voice, with a machine that shows its working.
THE REFERENCESreferences/SOURCES.md · THE-GATES.md · THE-METHOD.md
THE MACHINEscripts/golive.py · three commandstemplate — write the launch checklist to fill; check — GO or NO-GO for a plan (.md checklist or .json); gates — the eight gates and every item, with the evidence each needs. 27 checks in its selftest. Standard library only; nothing opens a network connection.
What it does once your AI has it.
template Write the launch checklist to fill
check GO or NO-GO for a plan (.md checklist or .json)
gates The eight gates and every item, with the evidence each needs
On the record An Example Co order-help agent's launch plan taken from NO-GO (5 blocking items) to GO, and 48 of 48 in the trigger test
What it refuses, and where it hands off.
It refuses ticking an item for the team or inventing its evidence; calling a NO-GO a GO; marking any gate but the human gate n/a; accepting a threshold written after the run it judges.
a market launch plan with positioning, pricing, announcements and a retro → THE PRODUCT LAUNCH CARD
a risk register, scoring risks or mapping which rules apply → THE AI RISK REGISTER CARD
bringing an AI assistant to a company's staff → THE ROLLOUT CARD
deciding which actions need a person's approval → THE AUTONOMY LADDER
whether a law applies, and security or compliance sign-off → the team's own lawyer and security lead
Checks what the plan says, not whether the kill switch or the alerts work; not legal, security or compliance sign-off. 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, law, method, machine, seals, refusals and hand-offs, ledger, score and what it carries · SKILL.md — the door your AI reads first · START-HERE.md — the first five minutes, for you
SOURCES.md + THE-GATES.md + THE-METHOD.md — in references/, in the card's own voice
scripts/golive.py — the machine: template · check · gates; standard library only, no network; selftest 27/27
WITNESS.md — the run on a realistic case as printed, and the trigger test on Haiku, Sonnet and Opus · SELFTEST.md — the machine's own test, as run at the cut · AUDIT.md, the twelve checks · MANIFEST.md, every file hashed
tests/evals.json — 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 · HOSTS.md, where it runs
system/short-prompt.md and system/system-prompt.md — the doors for any AI without skills
PROOF · SELFTEST 27/27 · WITNESSED ON A RUN · TRIGGER TEST 48 OF 48 ON ITS OWN PROMPTS · HAIKU 16/16 · SONNET 16/16 · OPUS 16/16 · 0 WRONG PICKS · EDITION 1 · PUNCH 2B4D63A24828A29E2834D7C1
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 962 — 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 AND SELFTEST.MD
golive.py, every command showing its working.
A small Python program. Your AI runs it; you can too. The blocks below are pasted from the card’s own record, as printed.
template
write the launch checklist to fill
check
GO or NO-GO for a plan (.md checklist or .json)
gates
the eight gates and every item, with the evidence each needs
selftest
27 checks that prove the machine; it must print 27/27
python3 scripts/golive.py template --kind agent --product "Example Co order-help agent" --launch 2026-11-09 --out empty-plan.md && head -14 empty-plan.md
wrote empty-plan.md · 25 items in 8 gates · agent
SOURCE · the card's gate list (references/THE-GATES.md): 25 items in 8 gates, the card's own launch standard
# Launch plan · Example Co order-help agent
product: Example Co order-help agent
kind: agent
launch: 2026-11-09
owner: (who says GO, by role)
Tick an item only when its evidence is written after it, as `key: value`. Dates are YYYY-MM-DD.
An item that truly does not apply is written `n/a` with `reason:`; only the human gate may be n/a.
## EVALS · evals passed on your own data at written thresholds
- [ ] EV1 The evals ran on your own data: real or realistic cases from this product's users, not only a public benchmark. — data: (which cases, how many)
- [ ] EV2 Pass thresholds were written down before the run (metric and number). — threshold: (metric >= number) · written: (YYYY-MM-DD)
- [ ] EV3 The last eval run meets every threshold, and it is dated. — result: (metric number) · ran: (YYYY-MM-DD)
- [ ] EV4 The evals ran on the build that ships: the same model, prompt, tools and settings. — build: (model, prompt and tools version)
GO-LIVE · Example Co order-help agent · agent · launch 2026-11-09 · plan plan-week-1.md
VERDICT · NO-GO · 16 of 21 MUST items met · 5 block
DECIDES · head of customer care
BLOCKING · every item that stands between this plan and GO
EV2 EVALS WRITTEN AFTER THE RUN the threshold was written 2026-10-30, after the run of 2026-10-28
KS2 ROLLBACK OPEN in the plan, not checked
LM2 LIMITS NO EVIDENCE checked, but no alert: write `alert: …`
MO3 WATCH MISSING not in the plan
HG2 HUMAN GATE MISSING not in the plan
WORTH DOING · these do not block
MO4 WATCH MISSING not in the plan
IN3 INCIDENT MISSING not in the plan
DA3 DATA MISSING not in the plan
DI2 DISCLOSURE MISSING not in the plan
GATES
EVALS 3/4 · EV2
ROLLBACK 3/4 · KS2
LIMITS 2/3 · LM2
WATCH 2/3 · MO3
HUMAN GATE 1/2 · HG2
INCIDENT 2/2 · clear
DATA 2/2 · clear
DISCLOSURE 1/1 · clear
MATCHED BY WORDS · lines with no ID, read as: DI1 (line 25)
RULE · GO only when every MUST item is checked with its evidence written beside it, every test (evals, rollback, kill switch, human gate) is dated within 30 days before the launch date, and every eval result meets its written threshold. SHOULD items never block; they are listed.
SOURCE · the card's gate list (references/THE-GATES.md): 25 items in 8 gates, the card's own launch standard
LIMIT · it reads what the plan says. It does not pull the kill switch, run the evals or watch the alerts fire; a GO is the plan's word, checked for completeness, dates and thresholds. It is not legal or security sign-off.
ASSUMPTION · a test counts when dated within 30 days before launch (the card's default; set --fresh-days)
THE SELFTEST, AS PRINTED
PASS the template holds 25 items in 8 gates, all unchecked
PASS the feature template drops the agents-only item (LM3): 24 items
PASS the empty template is NO-GO with all 21 MUST items named OPEN
PASS a complete plan is GO: 21 of 21 MUST, no SHOULD open
PASS the same plan without its kill-switch test is NO-GO naming KS2 only
PASS a checked spend cap with no alert is NO EVIDENCE, naming the missing key
PASS a kill-switch test 51 days before launch is STALE
PASS --fresh-days 60 accepts the same test
PASS a test dated after launch is flagged AFTER LAUNCH
PASS an eval result under its threshold is BELOW THRESHOLD, with both numbers
PASS an at-most threshold (harmful <= 0.01) is missed by 0.02 and met by 0.004
PASS percents compare as percents; mixed units are refused, not guessed
PASS a threshold written after the run it judges is refused
PASS a feature that takes no actions may mark the human gate n/a with a reason: GO on 20 MUST items
PASS the kill switch cannot be n/a
PASS an untagged line is matched by its words (a kill-switch drill is KS2, not KS1) and reported as matched
PASS a line no item matches is listed as unread, never counted
PASS evidence on indented sub-bullets belongs to the item above it
PASS no launch date is NO-GO: nothing can be dated against it
PASS a template placeholder is not evidence
PASS check from an empty folder: GO, exit 0, the same bytes twice
PASS the JSON form of the same plan gives the same verdict and writes --json
PASS --report writes the verdict with its rule, source and limit
PASS an unreadable plan exits 2 with a message, not a crash
PASS a missing plan file exits 2
PASS template then check: NO-GO with 21 blocking, exit 1
PASS gates prints every item with the rule, the source and the limit
selftest 27/27
THE TRIGGER TEST. The card's line was tried on Haiku, Sonnet and Opus beside 16 neighbouring skills, in a clean environment: 48/48 on its own prompts (Haiku 16/16 · Sonnet 16/16 · Opus 16/16), 0 wrong picks; no misses.
THE RUNS ARE PASTED FROM THE CARD’S OWN RECORD AS PRINTED. THE CASES ARE THE STUDIO’S OWN EXAMPLES (EXAMPLE CO IS FICTION); NO PRIVATE PERSON IS NAMED.
QUESTIONS AT THE TABLE
Asked before the wire.
What does it need from me?
Python 3; the launch plan as a Markdown checklist or JSON (the card writes the empty one).
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 selftest (27/27), a run on a realistic case with every command and verdict pasted into WITNESS.md as printed (An Example Co order-help agent's launch plan taken from NO-GO (5 blocking items) to GO, and 48 of 48 in the trigger test); and the trigger test, the card's line tried on Haiku, Sonnet and Opus beside 16 neighbouring skills: 48/48 on its own prompts, 0 wrong picks; the twelve-check audit (AUDIT.md). 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. Checks what the plan says, not whether the kill switch or the alerts work; not legal, security or compliance sign-off.
What will it never do?
It refuses ticking an item for the team or inventing its evidence; calling a NO-GO a GO; marking any gate but the human gate n/a; accepting a threshold written after the run it judges.
Who may use it?
One person, any number of their own AIs and computers; a team needs a company licence.
