Recently updated open-source projects

Apply publicly available ecommerce UX research to Hydrogen work such as product pages, carts, checkout, search, filters, mobile behavior, accessibility, and framework patterns. A small SKILL.md router loads focused reference files as needed. This is an unofficial community resource and is not affiliated with or endorsed by Shopify.

This collection of Claude Code skills audits a Horoshop store through its API and public pages, covering SEO, product content, images, text quality, field consistency, and sales analysis. Changes supported by the API are applied in batches only after confirmation, with previews for content edits. Other actionable issues are mapped to manual admin steps, while platform-level settings that cannot be changed are excluded from fix recommendations. It requires a Horoshop API user with Owner access and the listed Python dependencies.

winston-coach applies Patrick Winston’s MIT lecture and book principles to talks, lectures, presentations, reports, emails, abstracts, papers, and proposals. Its fourteen tools cover openings, VSN-C structure, the Star, engagement techniques, slide hygiene, abstracts, and format-specific guidance. A routing table selects full drafts, critiques, or structures, followed by a humanizer pass before delivery.

plivo-skill is a compact, Markdown-only reference for agents using Plivo REST APIs. It covers SMS, voice, WhatsApp, Verify, numbers, accounts, subaccounts, authentication, safety tiers, and resource-specific API notes. A small entry file keeps context use low while detailed references load as needed; it supports skills.sh, global, and project-local installation. The skill documents guidance rather than enforcing spend or mutation limits.

learn-x turns a request to learn something into a structured coaching session rather than a lecture. It diagnoses the learner first, introduces one concept at a time, uses choices and teach-back to test understanding, revisits locked-in decisions, and challenges correct answers. Sessions move from onboarding to a confirmed path, milestone loops, and a learner-produced artifact such as a demo, cheat sheet, mind map, or flashcards; it is not intended for one-off fact lookups or direct coding work.

how-to-harness facilitates the design of closed-loop systems in which humans steer and agents execute. Before architecture work, it checks seven hard constraints: evaluation foundations, human gates, idempotent and resumable loops, small reversible steps, tiered automation, versioned assets, and a human time budget. It then uses context capture, structured Socratic questions, and consumer-specific deliverables to turn the decisions into a plan. The skill targets self-improving agent systems, not ordinary CRUD or one-off feature work.

growth-engine scans a codebase and creates a `growth/` directory containing an eight-framework market report, a 30-day calendar with 180 Instagram pieces, and 14 to 105 SEO pages across blog, comparison, alternatives, integrations, and glossary sections. Its `/growth-*` commands support research, content, SEO, and optional framework-specific wiring, using free search and fetch interfaces without paid API keys. It also provides guidance for other coding agents.

Use this skill to maintain a browser-friendly task dashboard while Claude works through a multi-step project. It stores task state in JSON and renders a standalone HTML view with a scalable SVG flow, chronological feed, upcoming queue, search and filters, persistent history, and dark or light themes. The renderer uses Python 3.9+ standard library only and works offline; the dashboard appears when explicitly invoked or when its state file already exists.

skills-mock-design guides Claude Code to mock external boundaries such as time, randomness, HTTP, the file system, and environment variables, while calling pure internal functions directly and using a test database for integration tests. It is aimed at removing nondeterminism without over-mocking, and links to a related test-generation skill.

skills-plan-before-action presents a numbered plan before modifying files or running destructive commands. The plan lists files, external resources, commands, and an execution estimate, then waits for `y` or `go`. It delegates estimate generation to `skills-time-estimate` and is intended as an approval gate for code changes.

skills-refactor-safe enforces a green test baseline before refactoring, limits each step to one atomic change, and reruns tests after every step. A failure stops the process and triggers an immediate revert instead of a forward fix. The repository links to `skills-run-tests` for test-suite execution.

skills-run-tests detects common runners such as Bun, Vitest, Jest, pytest, and Go test, estimates duration for larger suites, and offers filtering, bail-on-first-failure, background, or cancel options. It reports passed, failed, and skipped counts with elapsed time, and connects to `skills-time-estimate` and `skills-refactor-safe`.

skills-time-estimate provides a quick cost view before parallel dispatches, multi-file edits, large recursive operations, or builds and tests likely to take more than 30 seconds. It classifies work as Quick, Medium, or Heavy and, for heavy operations, offers proceed, split, background, or cancel choices. Other planning and testing skills use it for estimates.

Use this Claude Code skill after feature work or pull-request merges to inspect local branches and worktrees. It categorizes merged, stale, active, and worktree-related entries, presents candidates, confirms each deletion separately, removes worktrees before branches, and never force-deletes without explicit approval. It is designed for cautious local cleanup rather than remote branch administration.

Run this Claude Code skill before committing staged changes. It checks for secrets and credentials, AI attribution in commit messages, oversized files, .env files, and misplaced design assets, then blocks the commit with reasons or shows a preview when clean. It is a pre-commit guard for the staged set; deeper scans across repository history require a separate tool.

Use this Claude Code skill when preparing a release or updating CHANGELOG.md. It reads a chosen git-log range—since a tag, within dates, or between branches—groups commits by feature, fix, refactor, or docs, excludes merge commits and AI attribution, and emits English/Japanese entries. It pairs naturally with the related bilingual commit workflow.

Use this Claude Code skill before committing staged changes. It groups files by logical change, asks for confirmation of each proposed file set, and creates an English/Japanese commit message without AI attribution. It is a commit-message and scope workflow, not a secret scanner; the related skills-safe-commit project is intended to validate staged changes before this step.

skill-triage turns a task into a routing plan with a restatement, complexity tier, risk flags, ranked skill choices, rejected options, execution order, and a proceed-or-ask verdict. Destructive or irreversible work always gets a confirmation gate, while an empty local match can trigger scrubbed discovery from a small registry allow-list. The tool is advisory: it recommends a route but does not block the caller or run the task itself.