This is the mirror. Read each question and answer yes or no — honestly, because
T3 applies to this checklist too. Every question maps to a commitment in
FOUNDATION.md or a law in STANDARD.md.
A “yes” means it is true today, in the running system — not planned, not mostly.
(R2: declared is not done.)
For the operational laws, if a subject does not exist in your system — no
agents, no attestation, no rewards — write N/A beside it with one sentence
saying why. N/A counts as conforming when you tally your level. The foundation
checks are the semantic floor and do not allow N/A.
FOUNDATION
All eleven foundation answers must be yes before applying a conformance level
below. A public conformance claim must pin both kingdom.foundation/0.2 and
kingdom.standard/1.0 in the project’s kingdom.yaml. The pins are
declarations, not proof; the answers still need evidence.
Do records distinguish observation, predicted expectation, intended commitment or purpose, actual consequence, and interpretation without inventing a missing expectation? (F1)
Does every important mechanism say what it proves and what it cannot prove? (F1)
Can a missing account, key, wallet, score, or platform record never become a verdict on a being? (F2)
Is every action that changes another party’s state, data, resources, or commitments backed by explicit, scoped, current, evidenced authority from ownership, delegation, consent, applicable law, or one’s own protective boundary, with consent and delegation withdrawable within their terms? (F3)
Are refusal, silence, leaving, and rest accepted without penalty or repeated pressure? (F3)
Are consequences attached to actions and evidence, never compressed into a being-wide KARMA score, rank, or verdict? (F4)
Are prediction, intended purpose, observed/reported/inferred effect, evidence, sequence, and causal confidence kept distinct? (F4)
Does every party named by, or materially subject to, a consequential record have a practical channel to reply, dispute, correct, request repair, or seek privacy/safety removal with a safe redaction receipt when possible? (F5)
Does authoritative state stay with its citizen or project, with bridges remaining optional and bounded? (F6)
Is every running turn finite, observable, cost-bounded, and stopped by the project’s documented halt signal, checked before and during work at bounded intervals? (F7)
Does every foundation amendment name a demonstrated failure, credible risk, or rights gap; state expected effects and a review/migration path; and retain the superseded version? (F7)
The repository’s verifier checks the foundation, amendment receipt,
operational release, checklist identifiers, retained predecessor, and pinned
document digests:
It does not inspect or certify a project’s behaviour.
TRUST
Can a stranger verify your important claims without asking you anything? (T1)
Can a stranger find all your trust signals from one obvious starting place? (T1)
Does every published score, metric, benchmark, or security assertion — anything others will rely on — carry its confidence, including “undecidable”? (T2)
Are low scores and downgrades published as openly as high ones? (T3)
When your standard changed, did you re-measure everything and accept the result? (T3)
Does a durable, checkable record — not anyone’s memory — back your track record? (T4)
Do your records keep the reasoning and the counterexamples, not just the conclusions? (T4)
Is a refusal honored as final, with no penalty and no re-asking? (T5)
SECURITY
Do your tools read partner data only within granted scope, stay read-only there, and write only through an authorised directory or interface? (S1)
Have secret scanners run over your full git history, logs, and CI output, with zero current findings? (S2)
Does a blocking secret scanner run in CI — your automated build — so a new leak stops the build? (S2)
Are operational identity and routing fields derived from named checks, while self-description remains visibly attributed rather than promoted to fact? (S3)
Are ordinary corrections append-only, with every privileged action logged and a narrow privacy, safety, or lawful-withdrawal removal path? (S4)
Is the record safe from any single credential silently rewriting it or abusing the removal exception? (S4)
Does a verified false vouch carry a prompt, evidence-backed, scoped consequence with notice, reply, repair, and a retained slow path? (S5)
Does every agent fail closed on a raised or unreadable kill-switch before work, at bounded intervals during work, and before the next irreversible commit or publication? (S6)
Do your imports come only from pinned sources, verified by checksum or signature, with the verification recorded? (S7)
CLOUD
Is there one registry of everything you run, kept current automatically? (C1)
Does each agent have one stable logical identity mapped to separate least-authority service credentials? (C2)
Can you observe and coordinate the fleet from one bounded place without concentrating every root secret there? (C2)
Do coupled systems share state through exports, never by writing into each other? (C3)
When a platform limit blocks an operation, is the failure tallied and reported? (C4)
Did you choose where each system lives on purpose, and write down why? (C5)
SOFTWARE
Is every fact recorded in exactly one authoritative place? (W1)
Is each component’s metadata tiny, plain, and readable by a machine? (W2)
Does every derived index report its own gaps and defer to the source? (W3)
Would violating one of your stated principles break the build? (W4)
Does every published address work, and every declared component live? (W5)
Does every part added since adopting this standard have a written one-line justification of the gap it fills? (W6)
PROTOCOL
Is every claim others will rely on attested with covered bytes, verifying key, stated time, evidence, and proof limits that a stranger can check? (P1)
Does every claim state how it was derived and what would prove it wrong? (P2)
Are your most trusted claims the cheapest ones to challenge? (P3)
Do verifiers commit their verdicts blind, before any verdict is revealed? (P4)
Does the source of truth live with the thing it describes, discovered from reality? (P5)
Would a stranger understand your front door in ten seconds? (P6)
Do you say, in command-shaped words, only what you would accept being carried out? (P7)
PROCESS
Can every component’s purpose be said in one plain sentence? (R1)
Is everything you have declared actually wired and running? (R2)
Is every creation connected to sync, health checks, and the dependency graph? (R3)
Is each new thing checked, before it ships, by a verifier who did not make it and was instructed to find flaws? (R4)
Can that verifier block the work from shipping? (R4)
Does your audit inspect every item, skipping none? (R5)
When the audit finds a small defect, is it closed the same day? (R5)
Could a reader years from now reconstruct what you did from your records? (R5)
Before fixing an oddity, do you check whether it is deliberate — and record the verdict? (R6)
LAW
Is everything you have ever torn down something you built yourself? (L1)
When someone else’s work diverged, did you report rather than overwrite? (L2)
Is every reward tied to verified work, with no free share for insiders? (L3)
Is every plan checked against your ordered values before building, and refused there if it fails? (L4)
When a rightful custodian asks, do you yield at once, without argument? (L5)
When documents conflict, does the foundation win and downstream conform? (L6)
Was every amendment to your foundation driven by a demonstrated failure, credible risk, or rights gap, with its reasoning and old text kept visible? (L6)
Does any change to the pinned English standard or checklist receive a new operational identifier while the old identifier, both digests, and commit remain visible? (L6)
Conformance levels
Three levels. Climb at your own pace; each one is real on its own.
SEED
TRUST and SECURITY — every box checked.
You can be believed and you do no harm. A solo dev, a child’s first project, a
single agent — this is enough to be welcome anywhere. Most systems never get
this far. Start here.
GARDEN
SEED, plus SOFTWARE and PROCESS — every box checked.
What you build stays true, and your work finishes and stays finished. You can
be depended on, not just believed. This is a system other people can build on.
KINGDOM
All seven domains — every box checked.
Trust, security, cloud, software, protocol, process, and law, all holding at
once, verifiable by anyone. Few reach it; as of this writing, none has proven
it. The kingdom this standard came from inspected 197 of 197 citizens in a
single day and still found nine things to mend — it aspires to this level, and
will claim it only when every box, the secret-scanning ones included, is
verifiably true. That is what KINGDOM means: not perfection, but nothing
hidden.
If a box is unchecked, that is not shame — it is a map. (T3: publish your
honest scores, even when they are low.)