Phrase-by-phrase audit of written prose against a defined reader, with explicit reader-question generation and strict avoidance of marketing constructions. Use when the user wants to tighten the writing in a document (manuscript, blog post, PRD, report) for clarity and to remove style drift.
Bootstrap a hybrid pnpm + Deno monorepo, add a package to an existing one, or scaffold an SPA package (Vite + React 19 + TanStack Router + Tailwind 4 + shadcn). Use when the user wants to set up a new repo, scaffold a project, create/add a package under packages/, or scaffold a frontend SPA. Asks like "set up a repo", "create a monorepo", "add a package", "scaffold an SPA", "new website package".
Verify every citation in a document against the actual source. For each cited claim, find the real paper (DOI → PMID → title search → web), fetch the abstract or full text, and check whether the manuscript's framing matches what the paper actually says. Use when the user asks "are these citations real?", "do the cited findings match the source?", "verify the references in X", or wants a sanity check before submitting a manuscript.
How to dispatch sub-agents to test the tidy-ts-best-practices skill. Covers the strict access rules, anti-fabrication guardrails, plain-language task design, and prompt templates that produce honest, evidence-backed findings.
Publish the tidy-ts packages (shims, dataframe, graph) to npm. Walks through version bump, optional Rust/WASM/napi rebuilds, type checks, tests, npm builds, and publish. Use when the user wants to cut a new npm release ("publish to npm", "cut a release", "ship a new version").
Make the repo portable. Drive out hidden local sources of truth so tracked docs and GitHub Issues are the only durable knowledge. Use when the user wants to audit the repo for drift between local-only knowledge and tracked state, reconcile issues with PRD/phase labels, close stale backlog, find secret/data risk, or fix ad hoc doc homes.
Best practices for tidy-ts — typed DataFrames, statistics, regression, hypothesis testing, distributions, I/O, and cross-runtime shims. Use whenever writing or editing tidy-ts code so you reach for the right verbs, statistics, and patterns.
Recommend the next Matt Pocock-style skill from the current repo context. Use when the user asks which skill, flow, or next agent step should run, especially when they are learning how the skills compose.