visualslisted

# design-doc-review:visuals Presentation-form check only — this skill does not evaluate whether the right topics are covered (see `design-doc-review:outline`) or whether prose is well-written (see `design-doc-review:readability`). It answers: **is this content trapped in a format that hides its structure from the reader?** **Target**: {{args}} — a file path, or a doc already in context. ## Framework Two independent detectors, both counting features of the actual text rather than judging quality: 1. **Diagram detector** — structural/behavioral content (topology, call order, state machines) encoded as linear prose is a format mismatch: prose is 1-D, the underlying information is 2-D (time × participant, or state × transition). Applies to the *Technical spec* layer of the Three-Layer Onion (how it works), not the *Problem* or decision-narrative layers — see guardrail on argumentative content below. 2. **Table detector** — N options scored against the same M criteria is a format mismatch when rendered as N separate paragraphs: the reader must hold every prior paragraph in memory to compare. Applies wherever the doc is enumerating parallel facts, most commonly *Alternatives Considered* / *Risks* / *Rollout stages*, not wherever it's making a values judgment. Both detectors are counting exercises: extract the axes from the prose and see if 3+ points exist on each. ## Checklist ### Diagram signals — check each candidate paragraph/section against these - [ ] **Sequence**: pa