| name | visual-explainer |
| description | Converts dense terminal output, benchmark data, and audit reports into interactive single-file HTML pages. Use when user says "visualize benchmarks", "generate diagram", "audit visual", "compound recap", or after benchmark-analyst, blueprint-reviewer, or compound-learner produce markdown output. Do NOT use for generating raw markdown reports, editing existing HTML files, or general web development tasks. |
Visual Explainer Skill
Converts dense AIOps terminal output and benchmark data into self-contained, interactive HTML pages. No build step required — all output is single-file HTML with CDN-only dependencies.
Commands
| Command | Input | Output |
|---|
/generate-web-diagram | Mermaid diagram spec or architecture description | Interactive diagram HTML |
/benchmark-visual | results/benchmark-report.md or raw JSON results | Sortable comparison table + bar chart HTML |
/audit-visual | Readiness audit markdown | Color-coded PASS/FAIL status grid HTML |
/compound-recap | Compound summary markdown | Visual compound summary HTML |
/fact-check | Blueprint review report markdown | Structured findings HTML |
Workflow
- Read the source — Read the input file (benchmark report, audit, compound summary, or review).
- Select template — Choose the appropriate template from
.claude/skills/visual-explainer/templates/.
- Populate the template — Replace all
<!-- DATA: ... --> placeholders with actual data. For Mermaid charts, generate the diagram spec from the data.
- Write the output — Save to
results/<output-name>-<date>.html alongside the source markdown.
- Open in browser — Run
open <path> (macOS) or xdg-open <path> (Linux) to display immediately.
Template selection
| Command | Template |
|---|
/benchmark-visual | templates/benchmark-comparison.html |
/generate-web-diagram or architecture | templates/architecture.html |
/audit-visual or /fact-check | templates/audit-report.html |
/compound-recap | templates/audit-report.html (adapted) |
Output naming
results/benchmark-visual-<YYYYMMDD>.html
results/audit-visual-<YYYYMMDD>.html
results/architecture-<name>-<YYYYMMDD>.html
results/compound-recap-<YYYYMMDD>.html
AIOps use cases
benchmark-analyst → /benchmark-visual
After benchmark-analyst writes results/benchmark-report.md:
- Extract TTFT (p50/p90/p99), ITL (p50/p90/p99), throughput, error rate per config
- Populate benchmark-comparison.html: sortable table rows = configs, columns = metrics
- Generate Mermaid bar chart for throughput comparison
- Include hardware config header (instance type, GPU count, model)
blueprint-reviewer → /fact-check
After blueprint-reviewer produces its findings:
- Map PASS/FAIL/PENDING items to audit-report.html status grid
- Color-code by severity (FAIL = red, PASS = green, PENDING = yellow)
- Render action items table with file paths and line numbers
compound-learner → /compound-recap
After compound-learner writes results/compound-<date>.md:
- Show elevated vs. kept-local lessons as a two-panel summary
- Highlight which steering files were updated
Design constraints
- All templates are self-contained (no external runtime, no build step)
- CDN dependencies: Mermaid.js (jsdelivr), Inter font (Google Fonts)
- Dark/light mode toggle using CSS custom properties
- Mobile-responsive layout
- Accessible color palette (WCAG AA contrast ratios)
Troubleshooting
For detailed error → cause → solution mappings, see references/troubleshooting.md.
Quick checks:
- Blank chart area: Validate mermaid syntax at mermaid.live; ensure x-axis labels and bar values have same count
- Raw placeholders in output: Grep for
<!-- DATA: in output file — all must be replaced
- Sort not working: Sort comparator must strip units before parsing numbers
- File won't open: Use
open on macOS, xdg-open on Linux; verify .html extension
Success Criteria
| Metric | Target | How to Measure |
|---|
| Skill triggers on visual/diagram requests | 90%+ | Test with 10 prompts: "visualize", "generate diagram", "audit visual", "benchmark chart" |
| Does NOT trigger on markdown report requests | 0% false triggers | Test with "write benchmark report", "create audit markdown" |
| All template placeholders replaced | 100% | Grep output for <!-- DATA: — zero matches |
| HTML renders correctly in browser | 100% | Open output, verify charts render, tables sort, dark/light toggle works |
| Output is self-contained | 100% | Disconnect network, reload HTML — all content except CDN fonts should work |