Index of every page on this site: llms.txt
All skills
Run it yourself

Build a one-page "technical note" in an academic-paper style — small-caps running head, big serif title with one highlighted phrase, abstract, a hero flow diagram on a lavender panel, numbered two-column sections, a booktabs table, and captioned figures — as a self-contained HTML file plus PDF. Use when the user asks for a technical note, a paper-style one-pager, "notes like this", or a visual explainer of a system, workflow, paper, tool, or concept.

  • 3 files
  • 27.5 KB
  • Updated 2 days ago
  • GitHub
Use this Skill: https://skilld.dev/gh/shauntheking/tech-note/tech-note

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈117 tokens for metadata: the name and description. ≈1.1k when used: this file.

Description uses 5.8% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈466 characters, or 5.8%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Tech note

Topic: $ARGUMENTS (if empty, ask what the note is about and who it's for).

The reference is assets/example.html — a complete note that uses every component. Read it fully before writing. Copy its <style> block verbatim; reuse its markup patterns.

Workflow

  1. Gather facts. About the user's code/system → read it. About an external topic → research primary sources. Never invent numbers, dates, or quotes. A chart without real data must say "schematic" in its axis label and caption.
  2. Outline in your head (fixed anatomy, one page):
    • Running head: Technical Note · <Field> + tiny mark on the right.
    • Title block: badge (SVG icon/monogram) + title with one .hl phrase + italic subtitle (one line, states the idea, not the topic). Thick+thin rule below.
    • Abstract: 4–6 sentences, Abstract. label. What it is, how it works, the key rule, the outcome.
    • Figure 1 (hero): the mechanism as a flow diagram on .panel. Left = actors/inputs, middle = the agent/engine, right = numbered steps + a decision diamond + keep/drop outcomes, loops drawn back with labelled arrows.
    • Two columns. Left: 1 Method (one paragraph) → Table 1 (4–5 rules, one .hl row for the key rule) → 2 Why …. Right: Figure 2 (chart/timeline) → Figure 3 (a)(b) small multiples → 3 <consequence>.
    • Every figure/table has a caption whose first sentence says what it shows and the rest says what to notice.
  3. Write ~/Claude/notes/<slug>/<slug>.html (create the folder). Self-contained: inline CSS and SVG, Google Fonts link only. <title> = the note title.
  4. Render and look: bash ~/.claude/skills/tech-note/scripts/render.sh <file.html> then Read the PNG and check it like a reviewer (see checklist). Exit code 3 = overflow: read <slug>.full.png, tighten, re-render. Repeat until pages=1 and the checklist passes.
  5. Deliver: open <slug>.pdf. Tell the user the HTML/PDF paths and what's in the note.

Diagram rules (inline SVG)

  • Plan coordinates on a grid before writing: hero viewBox ≈ 0 0 960 480; columns at x≈40 (actors, w 210), x≈300 (agent, w 150), x≈500 (steps, w 220), x≈760 (outcomes, w 170). Steps h=50 with 22px gaps; box text starts 14px from the left edge.
  • Use the CSS classes, not inline colours: groups .box .human .edit .keep .drop .dark .focus .note wrapping a <rect>; text .k (small-caps kicker) .t (serif) .big (bold mono) .m (mono) .d (description) .i (italic edge label) .c (centered); lines .ln (+.dash).
  • One .focus box only — the step the note is really about.
  • Arrows: marker-end="url(#ah)"; end 2px before the target edge. Route loops in empty gutters with orthogonal paths (M x y H x V y); never cross a box or a label.
  • Text must fit its box: mono ≈ 0.6 × font-size per char, serif ≈ 0.5 ×. Shorten labels rather than shrinking fonts below 11px.
  • A badge is an SVG icon or monogram. Use a photo only if the user supplies the image, and never a real person's likeness otherwise.

Review checklist (after every render)

  • pages=1; nothing clipped at the bottom; columns end at similar heights.
  • No text overflowing a box, overlapping a line, or colliding with another label.
  • Every arrow starts and ends on an edge; loop labels sit beside their line.
  • Exactly one highlighted title phrase, one .focus box, one highlighted table row.
  • Figure/table numbers are sequential and referenced in the order they appear.
  • Facts trace to sources; schematic charts say so.

Fitting one page (in this order)

Cut words first (abstract ≤ 5 lines, method ≤ 9 lines, table cells one line) → compact the small figures (shorter viewBox heights) → only then reduce spacing. Don't drop below body 13.5px / caption 12.5px. If it still overflows, the note has two ideas: split it.

Gotchas

  • Fonts load from Google; render.sh waits 6s. Offline → fallbacks (Georgia/Courier) shift widths, so re-check box text fit.
  • .page is fixed at 1000×1294 with overflow:hidden so the PDF is one page; that hides overflow in the PNG — trust render.sh's page count, not your eyes, for fit.
  • SVG <text> doesn't wrap. Multi-line labels = multiple <text> elements 16–18px apart.
  • Keep marker ids unique per SVG (ah, ah2, …).
  • The example's content is about the plan-and-delegate skill — replace all of it.

Source: SKILL.md on GitHub

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at a9e1a29. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 days ago

README badge

README badge for shauntheking/tech-note