AI skills for Coding agent

Manages BigQuery datasets, tables, jobs, and basic SQL analysis workflows.

Brainstorms and shapes OKRs that connect outcomes to measurable key results.

Forces design brainstorming before creative/feature work so agents explore approaches instead of jumping to code.

Drives a real browser to open pages, click flows, screenshot, and check console errors.

This skill is the starting point for building on Neon, whose Postgres comes bundled with auth, file storage, serverless functions, and an AI gateway. It covers setting up the Neon CLI and MCP server, a branch-first workflow, finding the right docs page, and choosing narrower Neon skills.

Writes Conventional Commit messages reduced to their essential intent.

Reviews code with Caveman evidence-first standards—tight diffs, proofs, no fluff.

Builds ChatGPT Apps SDK applications combining MCP servers with app UI and tools.

Writes and runs SQL against chDB (in-process ClickHouse for Python) for local analytics.

Guides correct use of the Claude API—models, tools, streaming, and common integration patterns.

Advises ClickHouse architecture choices—table engines, partitioning, and workload patterns.

Applies official ClickHouse schema, query, ingestion, and optimization best practices for agents.

Discovers and chooses the right Cloudflare products for apps, APIs, agents, storage, and security.

Designs and troubleshoots Cloudflare One Zero Trust and SASE deployments.

Reviews changes since a fixed git point along Standards and Correctness axes for real-engineering rigor.

Designs deep modules and shared vocabulary so package boundaries stay clear and maintainable.

Runs CodeQL static analysis with interprocedural dataflow and taint tracking across major languages.

This skill sets up and fixes Turborepo repositories. It covers turbo.json tasks, caching, Remote Cache, CLI usage, filters, environment variables, package boundaries, and CI. The SKILL.md is a short entry point that links to reference files.

Captures and reuses learnings across Cursor sessions so agents improve from corrections over time.

This skill lets an agent run a browser through playwright-cli commands so it can open pages, click, type, and read page snapshots. It can also take screenshots, handle tabs and storage, and help with Playwright tests. It suits web app checks and repeatable browser tasks.