mistraleuh/my-claude-skills

This is a collection of 15 slash command skills built for Claude Code, the official command line tool from Anthropic. The author groups them into roles: an architect, a bug hunter, a security auditor, a tech writer, a project manager. Each skill gives Claude a persona and a fixed workflow, so the assistant follows steps instead of jumping straight to code. The headline command is /do, which chains five other skills together in one shot. /todo writes a detailed ticket from a fuzzy request. /dev implements the ticket. /verify-dev audits the code line by line against the ticket. /build runs types, lint, and unit tests. /test runs functional tests. If any gate fails, /do loops back to /dev, up to three retries per gate, then either reports green or stops with a clear failure state. Installation is a git clone, then copy each subfolder into ~/.claude/skills/, then restart Claude Code. After that, plain English phrases like 'commit and push' or 'there is a bug on X' route to the right skill automatically. Each skill description lists trigger phrases, so the user does not need to memorize the slash names. The other skills cover quality work and documentation. /hunt reproduces and isolates bugs before claiming a fix. /security-review focuses on injection, auth, crypto, and dependency issues with CWE references. /refactor produces a line by line plan with risk notes. /knowledge generates a CONTEXT.md map of an existing repo, and /doc-update keeps existing docs in sync with code changes. /push writes the commit message and pushes. The author says the skills were built originally for a Next.js, FastAPI, and SQLite stack at his company, but the patterns are meant to be stack neutral. He reports roughly 80% fewer follow-up prompts and review cycles cut in half on production tickets. The license is MIT.