Skip to content

Skills

warmhub/warmhub-skills is a set of skills for Claude Code, Codex, and other AI coding agents. Each skill packages the WarmHub knowledge an agent needs to do one job well — designing a repo, planning an ingest, building a collector, wiring an app to WarmHub data — so the agent loads that context only when the task calls for it.

This page is the at-a-glance map of what’s published. For full install and update detail, see the skills repo README.

The skills assume familiarity with WarmHub’s core concepts — repos, shapes, things, and assertions.

Run from the project directory where you want the skills installed:

Terminal window
# Browse what's available
npx skills add warmhub/warmhub-skills --list
# Install everything
npx skills add warmhub/warmhub-skills --all
# Install one skill
npx skills add warmhub/warmhub-skills --skill build-warmhub-repo
# Target a specific agent (e.g. Codex)
npx skills add warmhub/warmhub-skills --all -a codex

To pull updates later, run npx skills update (aliases: check, upgrade) — it reports which of your installed skills differ from the published versions. See the repo’s CHANGELOG.md for what changed. There are no version numbers to pin: skills are versioned by content hash, not semver, so a skill is “out of date” exactly when its content differs from the published copy.

The skills are grouped by goal. If you’re not sure where to start, begin with warmhub-builder — it routes you to the right skill for your stage.

SkillWhat it doesUse it when
warmhub-builderCoordinates a WarmHub project from idea intake to the right builder stage.Starting a WarmHub project and unsure which builder skill comes next.
SkillWhat it doesUse it when
modeling-foundationsApplies WarmHub repo-modeling foundations — things vs. assertions, about-cardinality, and the four-direction traversability test.Grounding the data model before any design, build, or review work.
design-warmhub-repoDesigns or reviews a repo’s shapes, assertions, source identity, certainty, and derivation policy.Creating a new repo, modeling something as shapes, or judging whether a graph is ready to ship.
discover-warmhub-repoReads and summarizes an existing repo — shape inventory, counts, commit cadence, sensitivity signals — without mutating it.Inspecting an existing repo for a fact summary before building against it.
SkillWhat it doesUse it when
plan-warmhub-ingestionTurns a design summary or source into an ingestion plan covering access, transforms, idempotency, backfill, quality checks (QC), and write boundaries.Planning how a data source becomes WarmHub data, before writing the build.
build-warmhub-repoBuilds a complete ingestion repo — Bun/TS scaffold, shapes, source fetches, WarmHub operations, quality checks, auth, and a verified first ingest.Implementing an approved ingestion plan or repo design.
wh-commit-designDesigns commit plans for large or non-trivial ingests — chunking, sizing, resumable multi-commit workflows.Batching a large dataset into safe wh commit runs.
SkillWhat it doesUse it when
build-warmhub-collectorBuilds a mobile-first data collection app that submits end-user observations as things and assertions.Adding a write/ingress surface with forms, media, offline queueing, and provenance.
build-warmhub-displayBuilds a read-only display, report, notebook, or analysis surface for WarmHub data.Building an egress surface — dashboard, report, or notebook — over a repo.
connect-warmhub-appWires a TypeScript app or notebook to WarmHub — SDK setup, env config, server-side token handling, a one-fact connection probe.Connecting a display, collector, or app to WarmHub data safely.
SkillWhat it doesUse it when
add-warmhub-componentAdds or installs a WarmHub component — manifest design, subscription wiring, credential sets, install/doctor/update/teardown.Componentizing a repo or installing a published component such as Veritas.
veritas-designDesigns how a repo uses Veritas — which propositions get certainty opinions, who asserts, how to seed reputations, how to consume consensus.Wiring multiple sources to express belief-disbelief-uncertainty about the same binary propositions.
find-self-verifying-objectivesFinds adjacent, simpler optimization targets measurable from collected data without external ground truth.Optimizing any iterative or measurement-driven system without labels — applies beyond WarmHub repos.