Standing Mark
When the examination establishes warrant within its stated scope.
An independent intellectual press · New York
Your AI agent has an answer. Does it have the warrant to act?
A correct answer does not automatically authorize the action that follows.
When an agent proposes to move money, merge code, or submit a filing, the question is not only whether its answer is right. It is whether the evidence, conditions, and authority behind that answer support the proposed action. The Decision Integrity Protocol examines that warrant before execution. The result is a signed record of what the examination establishes — and what it does not.
Standing Assessment
When it does not: a signed record of what was examined and where warrant was not established.
SignedEd25519 · Canonical JSON
Independently verifiableOffline, against published keys
AccessMonthly Subscriptions · 1 USDC x402
What It Does
An examination before execution. A signed record after.
01
Submit the claim
Submit the proposed action and the record behind it: the evidence, qualifications, conditions, and authority on which it depends. The question is whether that record supports the transition from an answer to an action.
02
Examine the warrant
The service examines the submission against a named, published profile — a defined scope that states what the examination covers. The resulting record identifies what was checked, what was established, and the limits of the examination.
03
Receive the result
When warrant is established within the stated scope, the service issues a signed Standing Mark. When it is not, the service returns a signed Standing Assessment recording what was examined and where warrant was not established. The fee buys the examination. It never buys the outcome.
Demonstration
A recorded Standing Mark. Open to inspection.
Don't take the claim on trust. Inspect the record. Explore a recorded house self-test, including its result, examination scope, issuer, and complete signed record. The demonstration plays back the record as issued. It shows what a Standing Mark contains and how its stated scope travels with its result.
The complete signed record — the envelope — is published for independent inspection. Its signatures verify offline against the published keys. The Mark is signed by the service key, and the service key is certified by the publisher root.
Recorded Mark · Issued 7 October 2026
Record type
Standing Mark · WHP-STANDING-MARK-v1
Outcome
ESTABLISHED — source, context, and unknowns established under the admitted profile
Issuer
Wheeler Hubbell Publishing
Signature
Ed25519 · Canonical JSON · WHP-JCS-I1
Mark ID
WHP-SM-f9449b31a9fc74458dfde4696f68c041a242194113b3936503fe3362854a073f
The envelope
Published in full. Inspect it without an account.
What Is At Stake
The failure is not always a false answer. Sometimes it is an unauthorized transition.
These hypothetical examples show how a workflow can use the examination before carrying out an action.
Hypothetical · Capital movement
The escrow release
A treasury agent receives a request to release $75,000 because deliverables have supposedly met milestone specifications. The wording matches the contract. The underlying claim has not been verified. If the gate is integrated into the release path, the evidence behind the claim is examined before the transfer proceeds. An unwarranted transition does not proceed on that basis.
Matching the language of a contract is not the same as establishing that its conditions have been met.
Hypothetical · Engineering
The clean-looking merge
A coding agent declares a database migration backward-compatible. Its own qualification — that compatibility requires a read-replica pause — has disappeared from the conclusion. If the gate is integrated into the merge path, the claim's conditions are examined before the merge proceeds on that basis.
The issue is not whether the conclusion sounds technically credible. It is whether the condition that makes it credible survives the passage into execution.
Why This Exists
The dangerous answer isn't always the one that looks wrong.
An answer can be fluent, coherent, and factually correct — and still omit a governing condition, answer a different question, or exercise authority it was never given.
The problem becomes consequential when an answer stops being an answer and starts changing the world. A finding becomes a release instruction. A technical conclusion becomes permission to merge. A cited source becomes the basis of a filing.
Each transition asks something new of the claim. Evidence sufficient to support a description may not be sufficient to authorize an action. Decision Integrity examines that passage. It distinguishes what is established from what is inferred, what is supported from what remains unknown, and what a claim may establish from what it is permitted to authorize.
“Truth is not a tone of voice. It is not a reward conferred by confidence, age, status, repetition, fluency, fear, or force.”Wheeler Hubbell · Decision Integrity: The Continuous Work · Preamble — The Difference Between a Word and Its Force
The Named Failures
Four ways a claim can acquire force it has not earned.
These failures can occur even when individual statements sound credible. The critical question is what survives the passage from source to conclusion to action.
01
Coherent Substitution
The answer is coherent. It answers a materially different question.
Reasoning well about a materially different proposition and presenting the result as responsive.
02
Authority Leakage
Information acquires authority it has not earned.
An informational object or relation acquires more force than its record entitles it to exercise.
03
Object Drift
The words survive. The object changes.
The language remains stable while the thing being reasoned about changes.
04
Qualifier Stripping
A condition disappears. The conclusion changes with it.
When limiting language disappears, the surviving words can become more dangerous than a direct falsehood.
Definitions from Decision Integrity: The Continuous Work, Part IV — The Named Failures.
The Standing Mark
A signed record of how far a claim may be relied on.
A Standing Mark states how far a submission may be relied on under a named profile and stated limits. It is not a universal verdict on truth. It records the standing established by the examination: what the submission is entitled to support within the scope examined.
When warrant is not established, no Mark is issued. The service returns a signed Standing Assessment — a bounded record of what was examined and where warrant was not established. A negative result is information, not an absence of service. It preserves the boundary between what the record supports and what remains unwarranted.
“No claim acquires force merely by moving. Every material increase in force requires a basis that can be inspected.”Wheeler Hubbell · Decision Integrity: The Continuous Work · Preamble — The Difference Between a Word and Its Force
“Standing is what a thing is allowed to do under stated conditions.”Wheeler Hubbell · Decision Integrity: The Continuous Work · Part I — Standing
“If the test fails, do not falsely resolve the question. Deny only the unearned force.”Wheeler Hubbell · Decision Integrity: The Continuous Work · Part X — Test of Passage
Access
One protocol. Choose your evaluation volume.
Subscriptions are billed monthly for a defined evaluation allowance. Upon checkout, your sovereign API key and onboarding instructions are delivered immediately to your email for instant integration into GitHub Actions, CI/CD pipelines, or agent runtimes.
Developer Sentinel
Developer · 200 evaluations / mo
$149 / month
For individual engineers running automated agent workflows, pull request gates, and development testing.
Subscribe · $149/mo →Team CI/CD Gateway
Team · 1,000 evaluations / mo
$699 / month
For engineering teams examining proposed actions across production pull requests, deployment gates, and recurring agent workflows.
Subscribe · $699/mo →Institutional Sovereign
Institutional · 6,000 evaluations / mo
$3,500 / month
For organizations running sustained, high-volume decision-integrity examination across enterprise automated workflows.
Subscribe · $3,500/mo →How to submit evaluations
Once subscribed, you receive your sovereign API key (whp_live_...) via email. Plug it directly into your workflow:
1. In GitHub Actions (Standing Gate & Sentinel)
Add WHP_API_KEY to your repository secrets, then add this step to your workflow file (.github/workflows/sentinel.yml):
- name: Run Decision Integrity Gate
uses: wheelerhubbell/standing-gate@v0
env:
WHP_API_KEY: ${{ secrets.WHP_API_KEY }}
2. Direct REST API (cURL / Python / Node)
Submit claims to POST https://api.wheelerhubbellpublishing.com/v1/evaluate with your bearer token:
curl -X POST https://api.wheelerhubbellpublishing.com/v1/evaluate -H "Authorization: Bearer whp_live_..." -H "Content-Type: application/json" -d '{
"claim": "Deploy release to production",
"subject": "Production Release",
"provenance": { "source": "ci/cd pipeline", "authority": "release-engineering" }
}'
3. AI Agent MCP (Model Context Protocol)
Point Claude Desktop, Cursor, or your custom agent to https://api.wheelerhubbellpublishing.com/mcp and invoke tool whp_evaluate.
What counts as an evaluation?
One submission examined under the applicable supported profile, with one signed result where that service path issues one. Both outcomes count equally. The fee buys the examination, never a favorable determination.
Evaluation and billing details
Replaying the demonstration, downloading an existing record, and verifying its signature do not consume evaluations.
Included with every tier
- The submitted evaluation is charged whether its assessment is positive or negative
- A negative assessment is not issued as a WHP Standing Mark
- Retrieval and recovery do not trigger a second payment
- Payment does not supply missing source authority or warrant a favorable result
Autonomous Machine Payment (No Subscription Needed)
For autonomous AI agents paying per execution without human intervention: POST /v1/evaluations accepts 1 USDC on Base via the published whp-standing x402 exact-quote binding.
Inspect For Yourself
Examine the specification, the record, and the method.
The public specification describes the service. The capability record identifies the issuer and admitted profile. The foundational texts explain the discipline behind the examination.
OpenAPI Specification
The public API contract, machine-readable
Standing Capability Record
Signed issuer certificate and admitted profile
llms.txt
Machine-readable summary for agents
Why AI Guardrails Fail at Autonomous Execution
The foundational treatise
Coherent Substitution — Monograph
The named failure, examined in full
The Protocol as a Human Method
The eight governing questions, for institutions and people