claude-skills

Claude Skills Collection

A collection of skills for Claude Code and Codex that extend agent capabilities with specialized knowledge and workflows.

Related packs

Two tightly-scoped packs also live at cdd.dev/skills. Same install convention (npx skills add ckorhonen/<repo>):

• swe-skills — 15 swe: skills for engineering judgment work: PR risk review, repo introspection, performance and security audits, incident follow-up, ownership maps. Invoked in conversation.
• hone-skills — 8 hone: skills that run on a cadence (daily, weekly, per-PR) to fight code entropy: method brevity, naming clarity, duplication, broken-windows detection.

This repo is the broad collection — wider in scope, more varied in purpose.

About

Here are some skills I've found useful for various projects I've been building. Would love feedback if anyone else finds them useful—have at it! If you're interested in some of my writings on AI and software development, follow me at CDD.dev.

Available Skills

Skills are located in the skills/ directory.

Development & Testing

agent-browser

Headless browser automation using agent-browser CLI with element refs optimized for AI agents

• Scrape a product catalog with accessibility trees instead of CSS selectors
• Automate a multi-step checkout form submission
• Run parallel browser sessions to test race conditions in a web app

agent-engineering

Battle-tested engineering principles for AI coding agents — plan-first workflow, subagent delegation, self-improvement loops, and verification gates

• Set up a plan-first coding workflow that prevents agents from writing code before the design is approved
• Create a self-improvement loop that benchmarks agent performance and iterates on prompts
• Build a verification gate that runs tests before any commit is accepted

babysit-pr

Create, monitor, and shepherd a GitHub pull request through CI, review, merge, and best-effort deployment validation

• Open a PR that respects the repo's PR template and required reviewer assignments
• Watch CI after each push and summarize failures with fix suggestions
• Handle review comments from Greptile and re-request review after fixes are pushed

bird-fast

Interact with X/Twitter from the terminal using bird CLI with browser session authentication

• Post a thread announcing a new open-source release from a script
• Search for mentions of a product and summarize sentiment
• Schedule daily tweets from a content queue

ceo-review

Review documents and proposals through a CEO lens — technical depth, data-driven decisions, and long-term strategic vision

• Review a product spec for whether it aligns with the 3-year platform vision
• Critique an engineering proposal on ROI and competitive positioning
• Evaluate a go-to-market plan for defensibility and market timing

ck-review

Review documents and proposals through the lens of a senior principal engineer — clarity, pragmatism, and shipping velocity

• Review an architecture decision record for over-engineering and hidden complexity
• Critique a technical spec for ambiguous requirements that will block implementation
• Evaluate a proposed API design for consistency with existing conventions

codex-advisor

Get second opinions from OpenAI Codex CLI for plan reviews, code reviews, and hard problems

• Get a second opinion on an implementation plan before starting a large refactor
• Validate security assumptions in a cryptographic module
• Unblock a tricky bug you've been stuck on for 30+ minutes

coreml-optimizer

Optimize CoreML models for iOS and macOS deployment with quantization, palettization, pruning, and Neural Engine targeting

• Optimize a CoreML model for iPhone 15 Pro Neural Engine with 4-bit palettization
• Profile inference latency with Instruments and identify Neural Engine fallbacks
• Debug accuracy degradation after applying 8-bit quantization to a vision model

cto-review

Review through a CTO lens — platform thinking, mobile-first craft, and iterative execution