本文へ移動
cccskills
無料GitHub で公開

new-diagram

Scaffold a new TikZ diagram from the snippet gallery with prevention rules pre-applied (explicit node dimensions, coordinate map, directional edge labels). Compiles standalone, invokes tikz-reviewer with citations from tikz-measurement.md, and loops on revisions until APPROVED.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md7.0 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Create a New TikZ Diagram

Scaffold a diagram from templates/tikz-snippets/, check it against the prevention rules, compile standalone, run the reviewer with measurement citations, and loop until the diagram passes. Use this instead of writing TikZ from scratch; the snippets embed the invariants that tikz-prevention.md requires.

Inputs

  • $0 (optional) — snippet name (without .tex). One of the filenames in templates/tikz-snippets/. If omitted, list the gallery and ask the user to pick.
  • $1 (optional) — output path. Defaults to Figures/new_diagram.tex — ask the user if this target exists so we don't clobber.

Workflow

Step 1: Pick a snippet

ls -1 templates/tikz-snippets/*.tex

Current gallery (see templates/tikz-snippets/README.md for descriptions):

  • dag-basic — 3-node causal DAG (X → Y with confounder U)
  • dag-mediation — X → M → Y with direct path
  • did-two-period — two-period difference-in-differences
  • event-study — event-time coefficients with 95% CIs
  • timeline — horizontal timeline with staggered events
  • regression-scatter — scatter + OLS fit + confidence band
  • flowchart-3step — vertical flow with decision diamond
  • supply-demand — supply/demand with shifted demand

If $0 is not one of these, or is omitted, ask the user which to use.

Step 2: Copy the snippet to the output path

SRC="templates/tikz-snippets/$0.tex"
DST="${1:-Figures/new_diagram.tex}"

# Confirm before overwriting
if [ -f "$DST" ]; then
  echo "Output path already exists: $DST"
  # Ask the user whether to overwrite. Do NOT clobber silently.
fi

mkdir -p "$(dirname "$DST")"
cp "$SRC" "$DST"

Step 3: Edit the content to fit the user's intent

Ask the user what the diagram should show. Edit $DST with the Edit tool:

  1. Update the comment block at the top so the intent sentence matches the user's goal.
  2. Update the coordinate map comment if coordinates change.
  3. Rename nodes and edit labels. Keep node style (dag-node, flow-node, etc.) unless the meaning actually changes.
  4. Do not add a bare scale=X to the tikzpicture options. Coordinates shrink, text does not — collisions follow. Allowed forms: scale=X, every node/.style={scale=X} or scale=X, transform shape. See tikz-prevention.md P3 for the full rule and tikz-visual-quality.md for the scale=1.1 convention.
  5. Every new edge label must carry a directional keyword (above, below, left, right, above left, etc.). midway alone is a path position, not a direction — P4 violation.

Step 4: Prevention pre-check (MANDATORY)

Run the same shared Python checker /extract-tikz uses — this is the one tool that enforces both P3 and P4 consistently across the two skills:

python3 scripts/check-tikz-prevention.py "$DST"
  • Exit 0 → passed, continue.
  • Exit 1 → violations (stderr reports line, rule, snippet). Fix $DST and re-run until zero.
  • Exit 2 → usage error (missing file etc.).

Do NOT re-implement the grep inline. The Python checker correctly handles multi-line \begin{tikzpicture}[...] options and multi-line \draw ... node {...} spans that a line-oriented grep cannot see.

Step 5: Standalone compile

All snippets are \documentclass[border=4pt]{standalone} so they compile without a Beamer frame and without Preambles/header.tex:

cd "$(dirname "$DST")"
xelatex -interaction=nonstopmode "$(basename "$DST")" > /tmp/tikz-compile.log 2>&1

Check exit code and *.pdf file size. If compile fails, read /tmp/tikz-compile.log and fix the .tex source.

Step 6: Visual review via tikz-reviewer

Spawn the tikz-reviewer agent with Task (subagent_type=tikz-reviewer). Pass the .tex source and the compiled .pdf path. The reviewer is now required to cite the pass and formula from tikz-measurement.md for every CRITICAL/MAJOR finding — vague reports are rejected.

Loop:

  1. If APPROVED → go to Step 7.
  2. If NEEDS REVISION or REJECTED → apply fixes to $DST, re-run Step 4 (prevention pre-check), re-compile (Step 5), re-invoke reviewer.

Max 5 rounds. If after 5 rounds the reviewer is still reporting CRITICAL issues, surface the situation to the user — the snippet or the requested content may need redesign, not just tweaking.

Step 7: Optional — convert to SVG for Quarto

If the user plans to use the diagram in Quarto slides (not just Beamer), convert the compiled PDF to SVG:

pdf2svg "${DST%.tex}.pdf" "${DST%.tex}.svg" 1

Snippet-based diagrams are single-page, so a single .svg with the same basename as .tex is correct here. This differs from /extract-tikz, which produces a multi-page PDF from a Figures/LectureN/extract_tikz.tex master and names outputs tikz_exact_00.svg, tikz_exact_01.svg, ... (0-based filenames over 1-indexed PDF pages). If you need multi-page output for slide-by-slide reveals, use /extract-tikz instead.

Step 8: Clean up build artifacts

cd "$(dirname "$DST")"
rm -f *.aux *.log *.out *.synctex.gz

Leave the .pdf and .svg (if generated). They're what downstream tools use.

Step 9: Report

Print a summary:

  • Snippet used → output path
  • Reviewer verdict and number of rounds
  • .pdf size and page count
  • .svg path if generated
  • Reminder to \input or \includegraphics the diagram in the target Beamer/Quarto file

Why start from a snippet?

Writing TikZ from scratch reliably produces collisions because the author cannot visually estimate where curves and labels will land. The snippets embed the invariants that tikz-prevention.md requires — coordinate maps, explicit node dimensions, directional edge labels — so the diagram passes the prevention pre-check by construction. You can always deviate from the snippet; the rules still apply.

Cross-references

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Turn an incoming set of findings — from an AI reviewer, a referee report, a code review, a linter, or a second model — into verified fixes, without letting a confident misread damage correct work. Every finding is a CANDIDATE until checked against the actual source. Use whenever you receive review comments, audit findings, or a critique you did not write yourself, especially when the reviewer is a model or when the volume is too large to check by feel.

日本語の概要は準備中です。原文の説明を表示しています。

pedrohcgs/claude-code-my-workflow1,6572026年9月28日 更新

Enforce the replication-protocol.md rule by cross-checking numeric claims in a manuscript against the actual R / Stata / Python outputs. Report PASS/FAIL per claim against tolerance thresholds. Use before submission and before releasing a replication package.

日本語の概要は準備中です。原文の説明を表示しています。

pedrohcgs/claude-code-my-workflow1,6572026年9月28日 更新

Before and after changing anything shared — a function's return value, a signature, a schema, a label set, a config default, a constant, a file format — find every consumer and actually run them. Catches the change that looks purely additive but silently breaks a contract in a file you never opened. Use when editing shared code, adding a field/column/return element, renaming, changing units or defaults, or touching a pipeline that produces reported numbers.

日本語の概要は準備中です。原文の説明を表示しています。

pedrohcgs/claude-code-my-workflow1,6572026年9月28日 更新

Snapshot the computational environment for a replication package — detects the analysis stack (R / Stata / Python) and emits the right lockfiles (renv.lock + sessionInfo.txt, requirements.txt / environment.yml / uv.lock, Stata version + ado package list), records seeds and RNG kind, optionally writes a pinning Dockerfile, and produces a paste-ready "Computational requirements" block. Use when user says "capture the environment", "snapshot my dependencies", "pin the versions", "make a renv.lock / requirements.txt", "make this byte-reproducible", or before releasing a replication package to openICPSR / the AEA Data Editor.

日本語の概要は準備中です。原文の説明を表示しています。

pedrohcgs/claude-code-my-workflow1,6572026年9月28日 更新

challenge

無料

Stress-test a finding against the choices you did not make. Enumerates the discrete forks a competent analyst could have taken (measure definition, sample filter, control set, clustering level, weighting, functional form), runs the specification grid, and reports the distribution rather than a point estimate — then attacks the identifying assumption with named, computable sensitivity statistics. Use when the user says "is this robust", "challenge this result", "specification curve", "multiverse", "how sensitive is this", "what if I'd used a different measure", "stress-test my estimate", or before a result becomes a headline claim. NOT a reviewer of prose or code — it challenges the CLAIM.

日本語の概要は準備中です。原文の説明を表示しています。

pedrohcgs/claude-code-my-workflow1,6572026年9月28日 更新

Save a structured state snapshot before stopping or handing off. Captures the active plan, recent decisions, file pointers (with line numbers), open questions, and the next 1–3 actions into a checkpoint file under `quality_reports/checkpoints/`. Optionally proposes `[LEARN]` entries to add to MEMORY.md. Use when user says "checkpoint", "save state", "snapshot before I stop", "where am I", "wrap up the session for handoff", or before a long break / model switch / collaborator handoff. Companion to (NOT replacement for) the narrative session-log workflow.

日本語の概要は準備中です。原文の説明を表示しています。

pedrohcgs/claude-code-my-workflow1,6572026年9月28日 更新

pedrohcgs のスキルをすべて見る

このスキルの問題を報告する