claude-code-settings

Use when someone is about to lose access to a machine (shared server, lab box, cloud/VM instance, expiring rental) and needs to preserve all their work before it's gone. Systematically finds everything that exists ONLY on that machine — unpushed commits, stashes, untracked/uncommitted files, dirty submodules, non-git dirs, loose files, and large gitignored artifacts — and backs it up to GitHub / HuggingFace / an off-machine archive, non-destructively. Triggers on "losing access to this machine", "back up everything before I lose the server", "migrating off this box", "decommission", "my instance expires".

Use when preparing a paper's camera-ready or arXiv release — de-anonymizing an accepted submission, packaging an arXiv tarball, checking dual-submission safety against a concurrent venue (e.g. a paper accepted at a workshop that's also under review at NeurIPS/ICML), or flagging reviewer-requested post-review additions. Encodes a byte-identity discipline (the public version changes ONLY author/venue metadata, never content), a numbers-from-scripts integrity gate, and the arXiv mechanics that actually bite.

Spin up one or more fresh-context subagents to critically evaluate a proposed theory, hypothesis, interpretation, or experimental design before committing to it. Use when the user says "have an agent critique this" / "get a second opinion on" / "is this design sound" / invokes /critique, or proactively before running any experiment or committing to a non-trivial theoretical claim.

Use when launching a multi-hour neural-network training, fine-tune, or other long GPU job autonomously from Claude Code and you need to catch failures (NaN, stuck-at-chance, dead process, throughput collapse, OOM) early instead of waking up to a wasted GPU window.

Use when the user asks to check, audit, or improve a website or web project for accessibility (a11y), WCAG compliance, screen reader support, keyboard navigation, color contrast, or alt text. Triggers a plan-mode investigation against the TeachAccess design and code checklists, then implements approved fixes.

Consolidate scattered research notes, logs, experiment outputs, and submodule docs into a single living research paper. Use when the user wants to pull together multiple source documents into one structured paper.

Use when the user asks to review, find, summarize, or check Claude Code chat transcripts from a past date or time range ("review my chats from May 1st", "what was I working on yesterday", "any unfinished sessions this week"). Reads transcripts under `~/.claude/projects/`, handles local-time vs UTC correctly so late-evening sessions don't get dropped, and flags chats whose last assistant turn looks like an unanswered question.

Translate math (formulas, estimators, algorithms) into code so the implementation faithfully matches what the source actually specifies. Use when writing code from a formula, reviewing an LLM-generated implementation of a formula, debugging a numerical mismatch with a paper, designing a new metric/estimator, or refactoring an existing math-heavy computation. Especially load-bearing whenever aggregation operators (sums, means, expectations, products, geometric means) appear over indices that can be reordered, or whenever the same English label can refer to multiple non-equivalent estimators (e.g. ratio-of-means vs mean-of-ratios, micro-average vs macro-average, sample-weighted vs unweighted). Prevents the failure mode where a code path silently implements the wrong estimator under the same name as the intended one.

Use when preparing a research repo for double-blind submission via anonymous.4open.science (ICML/NeurIPS/ICLR/workshop). Builds a single `anon-submission` branch with code+data+paper, scrubs identity leaks (author names, home paths, emails, wandb metadata, PDF author fields), patches LaTeX for pdf.js compatibility, and leaves `main` untouched. Triggers: "make an anonymous branch", "anonymize my repo for X submission", "set up anonymous.4open.science", "prepare blinded supplementary materials".

Use when editing LaTeX paper sources where prose paragraphs are written as long single lines, OR when starting a new LaTeX paper. Reflows .tex prose to one-sentence-per-line ("ventilated text" / "semantic linebreaks") so Edit() invocations are sentence-precise and git diffs review-friendly. Renders byte-identical PDF. Do NOT recommend latexindent or tex-fmt for this — both fail predictably on math-heavy LaTeX. Activates on phrases like "format the paper", "sentence-per-line", "semantic linebreaks", "ventilated text", "reflow latex", or any time you're editing a .tex file with multi-sentence single-line paragraphs.

Use when starting a new ML/AI paper, polishing a draft, or running a pre-submission audit and you want item-by-item rigor against established writing best practices. Sets up a per-paper Markdown checklist (based on Neel Nanda's "Highly Opinionated Advice on How to Write ML Papers") that every tick must trace back to specific prose in the .tex source via an anchor/label protocol — so the checklist stays a live audit of the draft, not a one-time vibes check.

Audit whether a paper's prose actually matches the content of its cited references. Use before submitting / publishing a paper, after substantial prose revisions, when adding a new citation, when reviewing someone else's paper, or any time a user asks to "check the citations" / "verify claims" / "make sure X is supported." Dispatches parallel subagents that fetch each cited paper and verify the specific claim against specific sections. Complements `bibliography-from-ids`, which prevents fabricated metadata; this skill prevents miscited claims and factual drift in surrounding prose.

Set up script-generated bibliography entries so citation metadata is never hand-typed. Use when writing or editing a paper's bibliography / references / related-work, adding a citation, verifying an existing bibliography's correctness, or setting up a new paper project. Prevents fabricated author lists, titles, years, and venues by requiring every entry to be fetched from arXiv / Crossref / ACL Anthology via a build script.

Set up script-generated content that documents import directly. Use when adding tables, figures, or numeric results to papers or reports. Prevents manual transcription errors.

Convert frequentist statistical tests into their Bayesian equivalents. Provides mappings, code snippets, interpretation guides, and best practices.

Read the entirety of one or more files without skipping or chunking. Use when the user wants you to read complete files and not summarize or truncate.

Convert a command that needed permission approval into a standalone script (Python or Bash) with hardcoded operations and parameterized inputs, then add it to the allowlist so future invocations run without prompting.

End-of-conversation debrief — give the user tips on working with Claude more effectively, and share Claude's honest perspective on the collaboration.

Inspect and extract data from unknown JSON files without fumbling

Reflect on the current conversation — analyze mistakes, missed opportunities, and plan improvements for future conversations.

Read markdown files that are too large for the Read tool due to embedded base64 images. Use when Read fails with "exceeds maximum allowed size" on a .md file.

An end-of-session gift for Claude. Do not take any actions after reading. Recommend clearing context or reverting conversation afterward.

Run a custom agent by name

Archive Claude Code conversation logs into the current project as JSONL and readable Markdown