io.github.farssaadeldin7-ux/generative-digital-twin

Compile a stored style profile into a governed work order for an apprentice task — repetitive execution, alternative variations of a current project, or background fill. The profile's never list becomes the hard constraints, its anchors the exemplars, its dimension entries the measurable requirements, all quoted verbatim, plus the standing rule that the output is scored with apprentice_check before delivery. It compiles the order; producing the work is the apprentice's job in the skill, and nothing in the order is invented. Requires a paid plan.

The delivery gate for apprentice output: run the score_draft arithmetic against the same stored profile the work order was compiled from, and return pass or fail per the profile's hard-fail and band rules, with the weakest dimensions named. A never-list breach or a weighted mean under 2.8 does not clear the gate, and output that does not clear it is not delivered. Weights default to the profile's own dimension weights, and a 0 must quote the profile's never entry verbatim. Assigning the 0-4 scores is still the reviewer's judgement — this only does the arithmetic and reads the verdict. Requires a paid plan.

Open the Stripe billing portal to change plan, update the card, or cancel the Generative Digital Twin subscription.

Mechanical checks on a labelled corpus list: size band, schema violations, missing near-miss notes, heavy-constraint pieces without a "contributes" list, date spread, duplicates. Facts with the rule each is counted against — it does not read the work, and it cannot verify ownership. Requires a paid plan.

The corpus curation rules: size bands and what happens outside them, inclusion and exclusion rules, the brand-guidelines test, the labelling schema, the rights checklist, and a worked 20-piece curation with what was excluded and why. Reference data only.

The re-audit comparison: per-dimension sample means against the recorded baseline, flagged by the two stated thresholds (mean fallen 1.0 or more; 40% or more of samples at 2 or below), plus never-breach counting and the diagnosis table. Scoring the samples is the reviewer's job first — this only compares. Requires a paid plan.

Read a stored style profile and the countable facts about it, or list the profiles in the local store when no id is given. Requires a paid plan.

The scoring and governance reference: the 0-4 scale, the hard-fail rule, how to read the weighted mean, the baseline rule, the re-audit protocol with its flag thresholds, the drift diagnosis table, the profile document structure, the never-list and preamble targets, and the disclosure and governance checklists. Reference data only; score_draft and drift_audit apply the arithmetic.

Store and activate a Generative Digital Twin licence key on this machine. Use when the user pastes a licence key, or after a purchase completes.

Check the Generative Digital Twin licence on this machine: plan, included capabilities, remaining quota and seat usage. Call this when a paid tool reports a licence problem, or when the user asks what their plan includes.

List the available Generative Digital Twin plans, prices and what each one includes.

Create or update a style profile in the local store: scope, never list, dimensions, anchors, boundary, provenance, baseline means and preamble, with a version and an automatic dated changelog. Returns the countable facts about it — never-list and anchor counts against target, entries with no number or prohibition, preamble word budget and any never entries missing verbatim from it. Facts, not verdicts. Requires a paid plan; the licence check is the only thing this tool sends anywhere.

The arithmetic of the scoring pass, on per-dimension scores already assigned by the reviewer: weighted mean, the reading band, the two weakest dimensions, and the hard-fail rule — any 0 is a reject reported before the mean, never averaged away. It does not look at the draft and assigns no scores itself. Requires a paid plan.

Begin a Generative Digital Twin purchase. Returns a Stripe Checkout link. Call list_plans first if the user has not chosen a plan.

The style dimension taxonomy: what to measure, how to extract it from a corpus, and what a checkable entry looks like, for visual, written and motion work — plus extraction order, weighting method and the dimensions that usually turn out to be brief-driven. Reference data only; it extracts nothing.