mermaid-diagrams-applylisted

# Mermaid Diagrams Apply Audits and repairs a repository so its MkDocs Mermaid setup, its existing diagrams, and any newly authored diagram match the Mermaid Diagrams spec at `spec/project/mermaid-diagrams/<canonical_language>.md`. The skill both reports findings (read-only audit) and—with explicit per-item user consent—writes the missing or fixed pieces in place. ## Why this is a skill, not an agent - **Per-block user approval is the contract**: every Mermaid block emitted (or fixed) is shown to the user before it lands; the audit is read-only and the apply step is a sequence of approvals an agent's fire-and-forget shape can't carry. - **Output flows back into the main conversation**: the audit table, the per-finding fix proposals, and the diagram drafts surface inline so the user can redirect the type choice or the source pointer; isolating that in a structured-report boundary would obscure the per-block approval surface. - **Mid-flow type selection is load-bearing**: choosing between `flowchart` and `C4Component` for a portfolio map, or between `sequenceDiagram` and `flowchart` for a runtime workflow, is an interactive judgment call against the spec's catalog; that gating fits skill form, not an autonomous agent contract. - Counter-dimension considered: a narrower agent could specialize on parsing source artifacts (for example `pyproject.toml` → flowchart) and gain on context-window protection, but the high-impact part is the type-and-placement dialogue, not the parsing