Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Transfers prose edits from latex/index.tex (Overleaf) back into index.qmd. Use after pulling LaTeX edits from a collaborator.
.claude/skills/brycewang-stanford-sync-tex/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -29% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -11% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -8% | 0% |
Transfer prose changes made by collaborators in latex/index.tex (e.g., via Overleaf) back into index.qmd.
latex/.baseline.tex must exist (auto-created by bash scripts/render.sh)latex/index.tex must contain collaborator edits (pulled from Overleaf / GitHub)latex/.baseline.tex — the LaTeX snapshot from the last render (before edits)latex/index.tex — the LaTeX after collaborator editsindex.qmd — the current QMD manuscript sourcelatex/.baseline.tex vs latex/index.tex to identify changes.Focus on the document body only (everything between \begin{document} and \end{document}). Ignore the preamble (lines before \begin{document}) — these are auto-generated by Quarto.
\section{...}\label{...} and \subsection{...}\label{...} markers), check if the prose text changed between baseline and modified.index.qmd:\textbf{...} → **...**\textit{...} or \emph{...} → *...*\href{url}{text} → [text](url)Section~\ref{sec-id} → @sec-idFigure~\ref{fig-id} → @fig-idTable~\ref{tbl-id} → @tbl-id\cite{key} or \citep{key} → [@key]\citet{key} → @key\footnote{...} → ^[...]{{< embed >}} shortcodes exactly as they are in the QMD.Do NOT replace them with \includegraphics, \begin{figure}, \begin{longtable}, or any LaTeX equivalents. These shortcodes pull live outputs from notebooks and must remain intact.
\caption{...}), note the change in the summary report but do NOT modify the embed shortcodes — captions are defined in the notebook source cells.bash scripts/render.sh to verify the round-trip is clean (no warnings, no errors).latex/.baseline.tex does not exist, tell the user to run /project:render first.latex/.baseline.tex and latex/index.tex are identical, report "No changes detected" and stop.index.qmd, flag it and skip.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 4,492 | 12,021 | +168% | 1 | 1 | 0% | 208 | 1,079 | +419% | 0 | 0 | — |
case-02 | fail→fail | 5,844 | 5,591 | -4% | 1 | 1 | 0% | 251 | 1,043 | +316% | 0 | 0 | — |
case-03 | fail→fail | 4,591 | 4,908 | +7% | 1 | 1 | 0% | 197 | 1,194 | +506% | 0 | 0 | — |
case-04 | fail→fail | 13,117 | 6,762 | -48% | 1 | 1 | 0% | 2,186 | 1,057 | -52% | 0 | 0 | — |
case-05 | fail→pass | 11,373 | 1,759 | -85% | 1 | 1 | 0% | 1,735 | 1,073 | -38% | 0 | 0 | — |
case-06 | fail→pass | 9,665 | 5,239 | -46% | 1 | 1 | 0% | 1,664 | 1,685 | +1% | 0 | 0 | — |
case-07 | fail→pass | 12,788 | 4,226 | -67% | 1 | 1 | 0% | 2,191 | 1,556 | -29% | 0 | 0 | — |
case-08 | fail→pass | 10,286 | 4,452 | -57% | 1 | 1 | 0% | 1,755 | 1,559 | -11% | 0 | 0 | — |
case-09 | fail→pass | 11,527 | 5,878 | -49% | 1 | 1 | 0% | 2,039 | 1,870 | -8% | 0 | 0 | — |
case-10 | fail→pass | 8,809 | 2,989 | -66% | 1 | 1 | 0% | 1,786 | 1,406 | -21% | 0 | 0 | — |
case-11 | pass→pass | 3,761 | 1,948 | -48% | 1 | 1 | 0% | 790 | 1,107 | +40% | 0 | 0 | — |
case-12 | pass→pass | 5,087 | 2,905 | -43% | 1 | 1 | 0% | 900 | 1,344 | +49% | 0 | 0 | — |
case-13 | pass→pass | 6,320 | 2,423 | -62% | 1 | 1 | 0% | 1,179 | 1,173 | -1% | 0 | 0 | — |
case-14 | pass→pass | 5,522 | 2,072 | -62% | 1 | 1 | 0% | 981 | 1,133 | +15% | 0 | 0 | — |
case-15 | pass→pass | 4,071 | 2,159 | -47% | 1 | 1 | 0% | 662 | 1,165 | +76% | 0 | 0 | — |
case-16 | pass→pass | 6,730 | 3,344 | -50% | 1 | 1 | 0% | 1,152 | 1,347 | +17% | 0 | 0 | — |
case-17 | fail→pass | 2,891 | 1,557 | -46% | 1 | 1 | 0% | 494 | 1,032 | +109% | 0 | 0 | — |
case-18 | pass→fail | 15,231 | 3,685 | -76% | 1 | 1 | 0% | 2,440 | 1,374 | -44% | 0 | 0 | — |
case-19 | pass→pass | 11,462 | 3,590 | -69% | 1 | 1 | 0% | 1,942 | 1,457 | -25% | 0 | 0 | — |
case-20 | pass→pass | 6,464 | 4,165 | -36% | 1 | 1 | 0% | 1,015 | 1,487 | +47% | 0 | 0 | — |
case-21 | pass→pass | 7,198 | 6,428 | -11% | 1 | 1 | 0% | 1,377 | 1,657 | +20% | 0 | 0 | — |
case-22 | pass→pass | 8,090 | 6,607 | -18% | 1 | 1 | 0% | 1,463 | 1,809 | +24% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 18 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +27 percentage points is the difference between those two pass rates over the 18 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.