Skip to content
Portfolio/Capabilities

CAPABILITY 01 / 30

The guardians

Three checkers that read what was written and say exactly which rule it broke.

IN PRODUCTION

This runs in client systems today. The sample below was regenerated for a fictional company.

A brand system only holds if something checks it. These three run on the work itself: a brand guardian on markup and CSS, a phrase register on copy, and a price proofer on numbers. None of them rewrite a decision. They report, with the rule and the fix.

TRY IT / RUNS IN YOUR BROWSER

Use it here.

The mechanism is the production one. The company, the catalogue, and every figure are fictional.

Loading the three rule sets.

THE OUTPUT / GENERATED FOR APEX INSTRUMENTS

What comes out.

JSON
Apex brand rules, Evergreen system, version 1.4Data · Ten approved colours with their aliases and a 28 unit near miss tolerance, three approved font families with a 25 entry substitution table, and eight rules each carrying a severity, what it checks, why it exists, and who decides the fix.Open the file ↗
JSON
Apex phrase register, version 2.1Data · Forty banned phrases in four categories with 68 spelling variants and one regular expression pattern, each with the reason it was banned and one or two approved alternatives, plus fifteen approved phrases with the context each one suits.Open the file ↗
CSV
Apex price source, 823 rows, effective October 1, 2026Data · sku, item, category, list_price, partner_price, uom, effective_date. 256 Meridian configurations, 150 accessories, 160 consumables, 48 service plans, 57 calibration and training lines, and 152 spare parts. Every partner price is exactly ten percent below list. Nine rows are marked discontinued.Open the file ↗
MJS
The price source generatorData · Deterministic Node script that builds the CSV. Every price is a family base plus named option deltas, which is what keeps prices consistent inside a family. Run it from rebuild/ with node systems/guardians/make-price-source.mjs.Open the file ↗

THE MECHANISM / SHOW THE SHAPE, NOT THE SOURCE

How it works.

What a skill is
  1. 01Publish the rule

    A rule is written down once, with an id, a reason, and a fix.

  2. 02Read the work

    The checker parses the actual output, not a description of it.

  3. 03Fire precisely

    A finding names the value, the line, the rule id, and what to do.

  4. 04Fix the mechanical

    Colour snapping and phrase swaps are offered as one click. Judgement is left to a person.

  5. 05Report the rest

    Anything that needs a decision is escalated rather than guessed.

WHAT IT READS

  • The approved token set and layout rules
  • A phrase register of banned and approved wording
  • The price list the numbers must agree with

WHAT IT CHECKS BEFORE IT SHIPS

  • A near-miss colour is caught, not rounded away
  • Every finding cites a published rule
  • A price is never edited, only reported
  • A clean run still lists what it checked
In production: a brand guardian and a house-style scanner run across both company systems, and a price proofer checks an 823-row catalogue.

The mechanism runs inside client systems. Everything on this page was regenerated for Apex Instruments, a company that does not exist. Client names, source files, and internal wording stay private.

What it producesWhat is being built next