Five rules RedHub systems are built on

The five rules

1. 01 Worst signal first

   We don't average away the thing that can hurt you.

   An average can hide the one number that hurts you. RedHub's instruments name the weakest signal rather than averaging it away, and a roll-up across instruments takes the worst result, not the mean. The Agentic Executive Harness works this way: three green executives and one red is a red company.

2. 02 Gates over vibes

   A critical condition can stop the recommendation outright.

   Some conditions decide the outcome whatever the rest scores. Where a product defines a critical gate condition, that condition can worsen the result despite a strong average. The instruments listed below publish the gate and what trips it, so you can read the rule before you run it.

3. 03 No invented benchmarks

   Your inputs. Visible rules. Outputs you can trace.

   A graded instrument compares the inputs you supply to rules you can read, and the result is based solely on what you entered. RedHub does not independently verify the underlying facts.

4. 04 Read-only where it matters

   The system advises. You authorize.

   The system prepares the decision and you make it. The RedHub MCP connector follows the same rule: an AI assistant can list, explain and run the instruments you own, and it cannot buy anything.

5. 05 One fix first

   Where something is short, the verdict names the one constraint — not a list of twenty.

   Where something is short, the verdict names the one constraint to address next, not a list of twenty. RedHub treats a product that names a fix when nothing is short as a defect to correct, and a fix-first has to be able to move the verdict.

A good score is not a verdict

From the products’ own worked samples: each of these scored well on the average and was still decided by its gate.

• 100

  NOT CLOSEABLE

  AI Incident Postmortem & Readiness Gate

  Open-critical-action

• 100

  MISSING GUARDRAILS

  Voice Prompt Forge

  Safety-block gate

• 97

  WON'T TRAVEL

  Instagram Send-Rate & Share-Worthiness Lab

  Originality gate

How to read a RedHub verdict

A verdict is a named call, not a number

An instrument returns one of its own named verdicts. The score is how it got there; the verdict is what it says.

A gate can override the score

When a gate condition is met, the gate sets or caps the verdict whatever the average says. Run through the RedHub connector, the result says whether the gate fired.

AS DESCRIBED means “from what you entered”

Where a positive verdict may be shown to someone who did not supply the inputs, its name carries AS DESCRIBED. It describes your inputs, not a finding RedHub verified.

Not run is not passed

The Execution Layer and Ontology Layer roll-ups report an instrument they were not given as not run. They do not guess, and they do not count the gap as a pass.

The Standard, read from the products

44 instruments publish their rules in a form this page can read directly: the verdicts each returns, the gate that can override its score, and whether it names a fix first. It is not the whole catalog. Skills Packs, templates and playbooks are not graded instruments and are not listed.

Common questions