Tc Wiki

First Pass In A Repository

1. Find the wiki root, index.md, and log.md. Read them before changing structure.
2. Reuse the repo's existing folder layout and link style if one already exists.
3. If the repo has no established schema, use wiki-schema.md as the default.
4. Never modify raw source files.

Paper Frontmatter Access Status

For paper pages, add or refresh a source_access frontmatter field whenever you touch the page:

• full_text: the page is grounded in direct full article text or PDF access
• metadata_only: the page is grounded mainly in abstract, DOI landing page, bibliographic, or publisher-summary metadata
• citation_network: the page is grounded mainly in indirect citation traces or reference-list context
• mixed: the page combines more than one access mode and cannot be represented cleanly by one stronger category

Treat source_access as a maintenance signal, not as marketing. If access is constrained, say so plainly.

Core Operations

Ingest A Paper

1. Confirm the source is peer-reviewed or flag uncertainty.
2. Extract the minimum metadata needed for reuse:
   • title, authors, year, journal, DOI
   • basin, region, storms, or sample studied
   • data sources, methods, and variables
   • key findings, limitations, and uncertainty
3. Create or update the paper page first.
4. Add or refresh the paper page's source_access frontmatter field using the conventions above.
5. When a paper hinges on one or more central equations, preserve the most wiki-relevant equation(s) on the paper page in GitHub-compatible LaTeX math rendered by MathJax:
   • include only governing equations, diagnostic formulas, or compact budget relations that materially improve reuse
   • prefer one to three high-value equations, not full derivations or incidental algebra
   • define symbols in nearby prose and explain why the equation matters in the paper
   • treat the paper page as the canonical home for the equation; repeat it on topic or synthesis pages only when it becomes a cross-paper organizing concept
   • if source_access: full_text and the paper is equation-central, include a # Key Equations section
   • if direct full text is not available, do not reconstruct equations from memory or secondary summaries; record the gap explicitly instead
6. Add a # Referenced Papers section near the end of the paper page as a curated ingest queue:
   • list the most wiki-relevant cited papers, not the full bibliography
   • prioritize papers that support claims already propagated into topic, method, storm, dataset, basin, or synthesis pages
   • when a cited paper later gets its own wiki page, replace the plain citation with a direct wiki link
7. Update every materially affected topic, storm, dataset, method, and synthesis page.
8. Update index.md.
9. When present, refresh repository-wide indexes such as authors.md and tc-cases.md so they reflect the current ingest.
10. If README.md includes top-level wiki count links, check and refresh them on every ingestion using current repository totals:

• Papers (x): count current papers/*.md pages
• Authors (y): count current author entries in authors.md
• Storms (z): count current storm entries in tc-cases.md

11. Append a short chronological entry to log.md.

• Always add new entries at the bottom of the file so log.md remains in ascending chronological order (earlier entries first, later entries last).

Do not stop at the paper summary. The point of the wiki is to propagate new evidence into the existing synthesis.

Finalize Any Wiki Edit

Before finalizing any wiki mutation:

1. Run python3 scripts/lint_wiki.py from the repository root.
2. Resolve all reported errors.
3. If warnings remain, mention them explicitly to the user or record them in log.md when they materially affect confidence.
4. Ensure new evidence is integrated into the canonical page sections rather than appended as dated New Evidence-style blocks.
5. Ensure index.md blurbs describe current content, not edit history with labels such as new, seed, first-pass, or early.
6. Ensure every new page receives at least one intentional non-index inbound link from another content page.

Query The Wiki

1. Read index.md first.
2. Read the narrowest relevant pages before reading broader synthesis pages.
3. Answer from the compiled wiki, citing both wiki pages and underlying papers when possible.
4. If the result is durable, file it back into the wiki as a synthesis page, comparison, note, or table instead of leaving it only in chat.

Lint The Wiki

Look for:

• unsupported claims
• contradictions across pages
• single-study claims written as consensus
• stale syntheses not updated after newer papers
• repeated concepts or storms that still lack their own page
• orphan pages with weak linking
• terminology drift, especially for similar but non-identical metrics
• generated artifact drift in README.md, authors.md, and tc-cases.md
• schema drift such as missing required paper sections or ad hoc dated append sections
• access-status drift, especially inconsistent or missing source_access labeling on paper pages

Prefer using python3 scripts/lint_wiki.py for the repeatable structural checks, then do a human pass for scientific-content issues the script cannot judge.

Maintain authors.md

When the repository contains a top-level authors.md index:

1. Rebuild it from all current papers/*.md pages (not from memory or partial diffs).
2. Read each paper page's # Citation section and extract the listed authors as written.
3. Format every author as Last, First in the index.
4. Group entries alphabetically by letter with an A-Z jump list at the top.
   • Keep every A-Z section present; if a letter currently has no authors, leave the section blank (do not add placeholder text such as No authors indexed.).
5. For each author, list links to all wiki paper pages they co-authored.
6. Use abbreviated paper link labels:
   • one author: Surname (Year)
   • two authors: Surname1 and Surname2 (Year)
   • more than two authors: Surname1 et al. (Year)
7. Exclude placeholder/non-person tokens (for example, and coauthors) from authors.md; do not create entries such as Coauthors, Unspecified.
8. Run a duplicate pass before finalizing: detect likely duplicates where Last name + first initial match (for example, Fischer, M. S. vs Fischer, Michael S.), merge them into one canonical entry, and combine all paper links under that entry.
9. After updating authors.md, add or refresh its link from the top-level index.md page.

Maintain tc-cases.md

When the repository contains a top-level tc-cases.md index:

1. Rebuild it from all current wiki pages (not from memory or partial diffs).
2. Include only specific named historical/real tropical cyclones that are explicitly and materially discussed on current wiki pages.
3. Exclude idealized vortices, unnamed seasonal samples/composites, nature runs, generic basin summaries, and storms that appear only inside cited-reference titles unless the wiki text itself also discusses them as cases.
4. Group entries by basin and include a basin jump list at the top.
5. Order basin sections alphabetically.
6. Within each basin, create year subsections using storm year (not publication year) in ascending order.
7. Within each year subsection, sort storm names alphabetically.
8. Format each case entry as a bold storm name followed by two short sentences before the links:
   • first sentence: a compact storm description that states the maximum Saffir-Simpson category reached, the relevant basin/gulf/sea or track region, and whether the storm made landfall
   • second sentence: a short synthesis of why the case matters in the wiki
9. Use only links to current wiki pages; prefer the underlying paper page plus any topic/synthesis/method pages that explicitly discuss the named storm.
10. After updating tc-cases.md, add or refresh its link from the top-level index.md page.

Writing Rules

• Separate what the paper observed from what the paper inferred.
• Attribute claims precisely: Author et al. (Year) found ...
• Preserve disagreement. If papers conflict, state the conflict and possible reasons.
• Distinguish mechanism, evidence, and operational implication.
• Record limitations, especially small samples, retrospective framing, model dependency, basin specificity, and sensitivity to definitions.
• Do not invent numerical claims or thresholds the source did not make.
• Preserve only high-value equations that are central to the paper's method, diagnostic, or mechanism; do not transcribe full derivations or routine intermediate algebra.
• Use GitHub-compatible LaTeX math rendered by MathJax. Use inline $...$ for short in-sentence expressions and fenced math blocks for standalone equations.
• Transcribe equations faithfully and keep notation close to the source unless a small readability change is explicitly explained in nearby prose.
• If a useful point comes from interpreting a figure, mark it as an inference from the figure rather than a direct textual claim.
• If you rearrange, simplify, combine, or reinterpret an equation, label that as a wiki inference or restatement rather than a direct paper claim.
• Avoid duplicating the same equation across many pages; link back to the paper page unless the equation is central to a multi-paper synthesis.
• For review articles, keep cited-reference lists curated and functional. Do not paste the full bibliography into the main paper page unless the repository explicitly wants full bibliographies.
• Prefer updating a shared synthesis page over duplicating the same paragraph across many pages.
• Create a new page when a concept is central to multiple papers or is likely to become a repeated query target.

Hurricane-Specific Heuristics

When deciding which pages to touch, check across these buckets:

• storms and case studies
• basins and environmental regimes
• physical processes such as shear interaction, eyewall replacement, vortex tilt, air-sea fluxes, ocean cooling, rainband dynamics, surge, and waves
• metrics and predictors such as SST, OHC, TCHP, MPI, VWS, humidity, instability, and RI predictors
• observing systems such as aircraft, dropsondes, satellites, floats, buoys, and reanalyses
• model and analysis methods such as coupled models, idealized models, data assimilation, statistical tools, and machine learning
• synthesis topics such as rapid intensification, air-sea coupling, forecast error sources, hazard pathways, and climate relationships

Touch the narrowest pages that changed and at least one broader synthesis page when the new paper materially shifts understanding.

Evidence Standard

Prefer claims backed directly by the paper text, tables, figures, or supplement. Keep review articles separate from primary evidence. If a paper is outside scope, methodologically weak, or only partially relevant, say so plainly in the paper page instead of silently blending it into consensus language.

Suggested Session Pattern

1. Read index.md and the latest entries in log.md.
2. Read the target paper or the target wiki pages.
3. Draft updates in the paper page first.
4. Propagate those updates into topic, storm, dataset, method, and synthesis pages.
5. Update index.md and log.md.
6. Tell the user what changed, what remains uncertain, and what sources would most improve the wiki next.

Reference

Use wiki-schema.md only when the repository does not already have a clear wiki layout or when page conventions have drifted.