Worldbuilding Critic

Worldbuilding Critic — The System Judge

The worldbuilding domain has generators (magic-system-design, faction-design, extrapolation-engine) and a canon (world-bible), but no judge. This is it: is the invented system sound? It instantiates the shared discipline in _shared/critic-core (evidence-anchor cap, floor gate, reason-first, steelman, guards, bounded authority) with a worldbuilding-specific rubric. Read _shared/critic-core first — that machinery is inherited and not repeated here.

It answers a different question than consistency. "Does this contradict the axioms?" is the consistency gate (Layer 1). "Is the system itself good — costed, limited, generative, plausible?" is the soundness judge (Layer 2). A perfectly consistent world can still be decorative, un-costed, or one inference from a god-power.

How to Run

Input

1. The system(s) — the world-bible notes for a magic system / economy / society / technology (inline, paths, or a whole World-Bible/ + Concepts/ set).
2. Scope — system (one system) or world (the axiom set + how systems interlock). Default: system.
3. Precondition — the consistency audit (Layer 1) must pass first. Don't judge the soundness of a self-contradicting world.

Process

Layer 1 — Consistency audit (the hard gate; automatable). Before soundness, check the system doesn't contradict itself or the axioms. Extract entities, relations, and rules; look for logical, temporal, and causal contradictions across the notes (the corpus-inconsistency method: entity/relation + NLI + retrieval). Cite the conflicting passages. Any contradiction → FAIL Layer 1, stop, return the conflict. (This is the world's character-belief-tracker / world-bible consistency-audit analog.)

Layer 2 — Soundness rubric (the judge). Run the _shared/critic-core loop over the dimensions in references/soundness-rubric.md. Judge the function, cite the axiom or consequence, and apply the steelman before any FAIL (a deliberate genre choice — e.g. a soft, numinous magic à la Tolkien — is not a deficiency if it's owned and the story doesn't lean on it to resolve conflict). Run references/stress-tests.md — the "And Then What?" second-order detectors that catch the failures a forward-built world misses.

Aggregate by FLOOR GATE over the required dimensions (see the rubric). Emit a diagnosis, not a score: name the offending axiom / un-propagated consequence / un-costed power, quote it, and state the fix — for the world-builder or extrapolation-engine to act on. Withhold the tally (private rubric).

Output

SOUNDNESS VERDICT: PASS | FAIL   (floor gate · scope: <system|world>)
Consistency audit (Layer 1): PASS (precondition met)

| Dimension          | Verdict | Cited evidence (axiom / consequence)                    |
|--------------------|---------|---------------------------------------------------------|
| Cost is real       | PASS    | "every working consumes fuel-crystal... casting spends money" |
| Limits > powers    | PASS    | "acts only on living information... cannot raise the dead"  |
| Propagation        | PASS    | fuel=money → bankruptcy = visible decay (Axiom 1)       |
| Enforcement physics| FAIL    | "license... quasi-magically enforced" — mechanism unstated |
| ...                | ...     | ...                                                     |

Verdict legend: advisory dims marked FAIL* (non-gating); N/A = not applicable to this system type.

DIAGNOSIS (failed required dimensions only):
- <Dimension>: <offending axiom/gap>. Why it fails: <reason>. Fix: <propagate / cost / bound / stress-test, sourced to a generator skill>.

Scope Boundaries

Handles: judging the soundness of invented systems — cost, limits, access, propagation, plausibility, and the second-order stress-tests; with cited evidence and a diagnosis.

Does NOT:

• Generate systems — that's magic-system-design, faction-design, economic-systems, technology-progression (this judges their output; the diagnosis feeds back to them).
• Check prose quality — that's the quality-critic (a different domain).
• Adjudicate taste — soft vs hard magic, grim vs bright tone are the author's calls (bounded authority); the critic asks only "is it costed, limited, propagated, plausible?", never "is it the kind of world I'd build?"
• Decide canon — the world-bible is the authority; this flags soundness gaps for the human to rule on.

Related Skills

• Shared discipline: _shared/critic-core — the judging machinery this instantiates.
• Generators it judges (diagnosis feeds back): magic-system-design, faction-design, cultures-societies, economic-systems, technology-progression.
• Composes for evaluation: extrapolation-engine (the "And Then What?" engine — Layer 2's propagation + stress-tests).
• Sibling judges (peers, not dependencies — see _shared/critic-core): quality-critic (prose), whole-story-judge (macro narrative).
• Canon it checks against: world-bible.