About write-survey
write-survey is a research workflow skill for creating or maintaining a detailed LaTeX survey from a project's SOTA paper syntheses. It organizes papers by themes, integrates comparisons and research gaps, and validates coverage and builds.
Key features
- 01Create or update a LaTeX survey
- 02Group papers by themes and methods
- 03Integrate paper comparisons
- 04Identify gaps and research directions
- 05Track covered citekeys
- 06Build and validate the survey
Use cases
- 01Create a survey from digested papers
- 02Integrate newly added papers
- 03Remove excluded papers from a survey
- 04Check survey coverage and build output
Install
Copy the command and run it in a terminal.
npx skills add https:/ / github.com/ vincenzoimp/ academic-research-skills --skill write-surveyAfter installing, type/write-surveyin Claude Code to confirm it loaded.
Bundled filesreferences/survey-content.md
Details
From the project READMEWrite Survey
Produce or maintain survey/survey.tex: the single, complete, detailed
reading reference for the whole SOTA. After reading the survey, nobody
should need to reopen syntheses or PDFs for the large majority of questions.
Read First
survey/README.md— the survey contractsota/index.md— what the SOTA currently containssurvey/coverage.md— what the survey currently integratesreferences/survey-content.md— content checklist and style
The survey renders citations author-year (biblatex style=authoryear, per
survey/README.md) — a discursive review reads better than numbered [1].
Mode Selection
survey/coverage.mdlists no citekeys → create mode.- Otherwise → update mode. The user can force a full rewrite, which is create mode.
Create Mode
- Hard gate: read every synthesis. List all
digestedcitekeys fromsota/index.md, read everysota/papers/<citekey>/synthesis.md, and confirm the count matches before writing a single word of survey prose. - Design the grouping from what was read: themes, concepts, methodologies — whatever organization fits this SOTA best. Write the outline as LaTeX section comments first.
- Write section by section: per group, discuss each paper's contributions
and notable aspects in depth, compare approaches, name tensions and
contradictions. Respect each synthesis's "Safe claims / do-not-claim"
section. Cite with
\cite{<citekey>}only; every key resolves in the rootreferences.bib. - Write the mandatory final content section, Gaps and Research
Directions, grounded in the full picture per
references/survey-content.md. - Fill
survey/coverage.mdwith every integrated citekey. - Build (
make survey) and validate (make check). Fix all errors.
Update Mode
- Diff:
to_add= digested keys insota/index.mdmissing fromcoverage.md;to_remove= covered keys nowexcludedor gone. - Read the syntheses of
to_addand the survey sections they belong to. - Integrate additions into the right groups — extend or restructure a group when the new papers change its internal logic; never just append.
- Excise removals: delete their discussion, repair comparisons and transitions that referenced them, re-check the gaps section.
- Update
coverage.md, rebuild, runmake check.
Rules
- Every statement about a paper is supported by its synthesis (or its PDF when more depth is needed) — never by model memory.
- The survey is self-contained: detailed enough to replace re-reading the SOTA for most purposes. No length limit; completeness wins.
- Never edit
references.bibfrom this skill: a missing entry means the SOTA work is incomplete — stop and report.
Done When
coverage.mdequals the set of digested citekeys (or the user-approved subset),survey.pdfis rebuilt,make checkpasses