AI Agent Workflow Collection

Create clear diagrams and visualizations for documentation, technical reports, and data analysis. Prefer D2 (with ELK layout) for architecture diagrams. Use ASCII art for plain-text docs and terminal output. Use Mermaid only as a last resort. Also covers Python charts (histograms, box plots, heat maps, 3D surfaces, scatter plots, flame graphs) for data visualization. Companion to the handson-workshop skill.

Create a comprehensive, self-contained hands-on workshop for any technical topic. Use when the user invokes /handson-workshop (e.g., /handson-workshop "git internals", /handson-workshop "HTTP caching", /handson-workshop sqlite). First interviews the user to gauge audience knowledge-level and desired depth, including a short batch of technical multiple-choice questions that calibrates depth beyond self-report, then explores the local system to confirm the tooling needed to reproduce the experiments (offering to install missing prerequisites with the user's approval), and may suggest narrowing an over-broad topic into a smaller starter workshop plus a follow-on series. Then it produces (1) WORKSHOP.md with modules — each containing a best-fit diagram (block, sequence, flowchart, state machine, venn, or a bar/histogram/box chart), sample code for technical topics, hands-on commands with real captured output, and a quiz that links to an answer key — (2) WORKSHOP.html, a styled, self-contained HTML version with sidebar ToC, collapsible answer keys, and syntax-highlighted code, and (3) a standalone exam.html with 15-20 auto-graded questions (1-100 score, 70 to pass). Workshops are grounded in concrete output, not invented examples. For large or multi-module workshops it can instead emit a multi-page "wiki" (a landing page plus one page per module, with a persistent cross-page sidebar and prev/next navigation) via a separate generator.