mizchi/skills

The review-image skill allows you to review screenshots or other images using OpenRouter vision models via bundled Deno scripts. It is designed for:

• Quick Visual Regression Testing (VRT) sanity checks.
• Screening for invalid images.
• Implementing CI gates.

There are two primary execution modes:

1. scripts/review-image.ts: Returns freeform feedback.
2. scripts/review-image-ci.ts: Returns strict pass|fail JSON and exits with a non-zero exit code on failure, making it suitable for automated CI pipelines.

The sqlc-gen-moonbit-safety skill provides post-generation safety checks for environments using sqlc-gen-moonbit combined with Cloudflare D1.

It is specifically designed to gate against two issues:

1. BigInt-bind hangs: Preventing hangs related to D1 error code 1101.
2. SQL placeholder mix: Preventing issues caused by mixing different SQL placeholder styles.

The frontend-review-performance skill is designed for reviewing React rendering performance. It provides guidance and diagnosis for:

• Profiler-first diagnosis
• Correctness of memo, useCallback, and useMemo
• Virtual scroll implementation
• Use of useTransition and useDeferredValue
• Separation of Canvas/WebGL for data-heavy UIs

This skill covers the concepts outlined in checklist 24-rendering-performance.md.

The playwright-test skill provides best practices and references for Playwright Test (E2E). It is designed to assist when writing, reviewing, or configuring CI for Playwright tests.

Key coverage areas include:

• Writing effective tests
• Avoiding fixed waits
• Network triggers
• Drag and Drop (DnD) implementation
• Shard/retry setup on GitHub Actions

The actions-ci-tuning skill is designed to audit and improve GitHub Actions workflows. It provides expertise in:

• Cache setup for npm, pnpm, and yarn.
• Optimizing job parallelism.
• Implementing shard-based test splitting.
• Avoiding artifact handling pitfalls.
• Applying Playwright-specific patterns.

When to trigger this skill:

• When CI is slow.
• When encountering cache misses.
• When dealing with flaky shard jobs.
• When merge-reports fails.
• Upon an explicit 'tune CI' request.

The frontend-review-security skill is designed for conducting frontend security reviews. It covers three main areas:

1. Static Analysis: Identifying risky HTML patterns and environment variable exposure.
2. Authentication/Authorization Audit: Auditing token storage, route guards, and logout implementations.
3. AI Self-Penetration Testing.

The skill executes via the scripts/audit-security.sh script.

Note: If you need to perform CVE triage or detect deprecated libraries, use the frontend-review-deps skill instead.

Cloudflare Images is an end-to-end image management solution that provides storage, transformation, optimization, and delivery at scale via Cloudflare's global network.

Key features include:

• Automatic Optimization: Automatic AVIF/WebP format negotiation.
• On-the-fly Transforms: Resize, crop, blur, and sharpen via URL or API.
• Workers Binding: Transform images directly within Cloudflare Workers (the primary method for 2026).
• Direct Upload: Secure client-side uploads that bypass the need for a backend proxy.
• Global Delivery: Images are cached across 300+ Cloudflare data centers.
• Watermarking: Programmatic image overlays.

The nix-setup skill provides templates for setting up reproducible development environments using devbox (Nix-backed) or pure Nix flakes.

It includes preloaded configurations for:

• MoonBit
• Rust
• TypeScript + pnpm
• Python + uv
• Haskell
• OCaml
• OxCaml

Supported configuration files and integrations include:

• devbox.json
• buildNpmPackage
• direnv
• GitHub Actions
• Bootstrapping in sandboxed environments (e.g., Claude Code web)

Use this skill when starting, adding, or troubleshooting a Nix or devbox setup.

Cloudflare Network Interconnect (CNI) provides private, high-performance connectivity to the Cloudflare network. This service is available to Enterprise-only customers.

Connection Types

• Direct: Physical fiber in a shared datacenter (10/100 Gbps). Requires ordering a cross-connect.
• Partner: Virtual connectivity via partner SDN providers such as Console Connect, Equinix, or Megaport.
• Cloud: Connectivity via AWS Direct Connect or GCP Cloud Interconnect (available via Magic WAN only).

Dataplane Versions

• v1 (Classic): Supports GRE tunnels, VLAN, BFD, LACP, and peering. Note that MTU is asymmetric (1500↓/1476↑).
• v2 (Beta): Does not support GRE or VLAN/BFD/LACP (uses ECMP instead). Supports a consistent 1500 MTU in both directions.

The extract-glossary skill is used to extract and generate domain-specific terminology, industry jargon, internal/product terms, repository implementation maps, technical configurations, and Mermaid diagrams for onboarding from a specified repository, multiple repositories, or a GitHub organization.

Trigger this skill when a user requests tasks such as:

• Creating a glossary or domain dictionary.
• Generating onboarding materials.
• Summarizing technical terms from a repository or organization.
• Building a knowledge base that an AI can use without needing to re-verify information.

sql-lint provides static linting for sqlc-style SQL catalogs. It is designed to catch common mistakes such as:

• Duplicate query names
• Missing semicolons
• Use of SELECT *
• Double-wildcard LIKE patterns
• Other common, easily missed SQL errors.

It also supports optional sqlfluff integration for comprehensive SQL style coverage.