Code Walkthrough

Code Walkthrough — Source Code Explanation HTML Generation Skill

Split source code from the top into processing units (segments), and generate a single self-contained interactive HTML that pairs the fully embedded code with detailed commentary, SVG diagrams, and a glossary panel.

For progressive loading, this skill splits the detailed spec into reference/. As shown in the table below, read the relevant file at the point you need it.

Usage

Arguments: source file paths (required, multiple allowed), output HTML path (optional). If the last argument ends in .html, treat it as the output destination.

Single file

Pass one source file. If an output destination is specified, generate there; if omitted, generate walkthrough-<filename>.html in the same folder as the source.

Multiple files (combined into a single HTML with tabs)

Passing two or more source files switches to multi-file mode, combining them into a single HTML with a tab-switching UI.

• When the output destination is omitted, generate walkthrough-<a>-<b>[-...].html in the folder of the first file. If the combined name exceeds 80 characters, fall back to walkthrough-multi-<N>files.html
• Each source file = one tab. The tab label is the base file name (without extension). When names collide, such as main.py, prefix the parent folder name (e.g., 08_rlm_real/main)
• If even one file does not exist, exit with an error

Execution steps

1. Parse the arguments:
   • Just one file → single-file mode
   • Two or more files → multi-file mode (see "Multi-file tab spec")
   • If the last argument ends in .html, separate it out as the output destination
2. Confirm that all source files exist. If even one is missing, exit with an error.
3. Determine the output destination:
   • If specified by argument, that path
   • Single file: walkthrough-<filename>.html (same folder as the source)
   • Multiple files: walkthrough-<a>-<b>[-...].html (folder of the first file; if too long, walkthrough-multi-<N>files.html)
4. Read each source file in full.
5. For each file, write the commentary following "Segment splitting rules" and "Structure of each segment". In environments where web search is available, when a library, API, or spec you use is new / your knowledge is stale / you are unsure, check the official docs before writing the commentary.
   • Identify "technical terms that even a general engineer would find hard to understand without specialized knowledge" appearing in the commentary, underline them with <span class="term"> following reference/glossary-panel.md, and register their overviews in GLOSSARY.
   • When you diagram a processing flow with SVG, run the "SVG verification phase" in reference/svg.md every time you make a diagram, and confirm that all arrows connect to their targets before moving on to the next segment.
6. Assemble and output the HTML, starting from the bundled template:
   • Copy assets/template.html to the output path and fill its placeholders following reference/html-layout.md. The CSS/JS (layout, interactive features, glossary panel, tab switching) is already implemented in the template — do not rewrite it
   • Multiple files: additionally follow reference/multi-file-tabs.md
   • Register the glossary terms following reference/glossary-panel.md
   • When the output exceeds what can be written out at once (files over 500 lines / a large combined total across multiple files): generate and run gen_walkthrough.py under a temporary directory (/tmp, $TMPDIR, or on Windows %TEMP%, whichever path fits the environment) that fills the template and writes the HTML
7. Open it in the browser with "Browser display command".

Reference (read when needed)

The detailed spec is split into reference/. Read the relevant file at the point it becomes needed within the execution steps.

The template assets/template.html implements the page skeleton, CSS, and all interactive JS; the generated HTML must start from it rather than hand-written CSS/JS.

Browser display command

After generating, open it in the default browser matching the OS of the execution environment ($OUTPUT_PATH is the path to the output HTML):

• macOS: open "$OUTPUT_PATH"
• Linux: xdg-open "$OUTPUT_PATH"
• WSL: cmd.exe /c start "" "$(wslpath -w "$OUTPUT_PATH")"
• Windows (cmd): start "" "%OUTPUT_PATH%"
• Windows (PowerShell): Start-Process "$OUTPUT_PATH"

To auto-detect in a POSIX shell:

if grep -qi microsoft /proc/version 2>/dev/null; then
  cmd.exe /c start "" "$(wslpath -w "$OUTPUT_PATH")" 2>/dev/null   # WSL
elif command -v open >/dev/null 2>&1; then
  open "$OUTPUT_PATH"                                              # macOS
elif command -v xdg-open >/dev/null 2>&1; then
  xdg-open "$OUTPUT_PATH"                                          # Linux
fi

Pre-completion check

Before outputting the HTML, review the "Checklist" at the end of each reference file corresponding to the features you used. At a minimum, always review reference/segments.md (code embedding, commentary), reference/html-layout.md (layout), and reference/glossary-panel.md (glossary panel). If you used SVG or multiple files, also review the checklists of reference/svg.md (that the SVG verification phase has been run) and reference/multi-file-tabs.md respectively.