Skill Claude CodeCodex
Use when untrusted input may reach SQL, HTML, filesystem, shell, redirects, or other dangerous sinks. Covers validation, normalization, parameterization, output encoding, and refusing vague “sanitize everything” advice.
Reusable AI skill packs organized by library and topic.
This repository also configures its own agents. See what agent-skills tells them →
Skill Claude CodeCodex
Use when untrusted input may reach SQL, HTML, filesystem, shell, redirects, or other dangerous sinks. Covers validation, normalization, parameterization, output encoding, and refusing vague “sanitize everything” advice.
Skill Claude CodeCodex
Use when initializing shadcn, managing components.json, adding components or blocks through the CLI, or deciding how registries and updates should work without losing local ownership. Covers CLI defaults, registry boundaries, aliases, CSS entrypoints, and safe update posture.
Skill Claude CodeCodex
Use when composing shadcn components into real product surfaces instead of treating them like isolated snippets. Covers asChild, trigger ownership, dialog and dropdown composition, local wrappers, and the rule that the value is in stable composition patterns, not in blindly stacking primitives.
Skill Claude CodeCodex
Use when building forms with shadcn in a way that stays opinionated instead of devolving into copied snippets. Covers field primitives, choosing one form-state strategy, integrating validation, and keeping complex stateful form logic out of the visual component layer.
Skill Claude CodeCodex
Choose whether shadcn/ui is the right fit and adopt it with the right ownership model. Use when tasks mention shadcn setup, copying components into the app, owning UI source code, deciding between shadcn and a packaged component library, or routing work to theming, registry, composition, or forms follow-up skills.
Skill Claude CodeCodex
Use when defining shadcn theme defaults that the model should not improvise per component. Covers CSS variables, semantic tokens, OKLCH-based palettes, dark mode, and the rule that visual consistency should come from tokens and composition instead of one-off utility overrides.
Skill Claude CodeCodex
Use when creating or refining an AI skill. Covers trigger wording, scope boundaries, progressive disclosure, references versus scripts, and keeping SKILL.md concise and installable.
Skill Claude CodeCodex
Use when modeling products and prices or implementing Stripe Checkout with sane defaults. Covers products, prices, lookup keys, Checkout Sessions, hosted checkout as the default path, and the boundary between checkout UX and webhook-based fulfillment.
Skill Claude CodeCodex
Use when implementing Stripe's customer portal as the default self-serve billing surface. Covers portal configuration, product-catalog boundaries for upgrades and downgrades, portal session creation, deep links, and webhook-driven sync after customers manage their own billing.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: