Claude Skills

Claude Skills

A marketplace of Claude Code plugins built on one idea: AI is fast; these tools make the parts it likes to skip hard to skip. The gates enforce your standards, the guides keep the AI working from current best practice instead of stale training data, and you make every decision. The reasoning behind that is in PHILOSOPHY.md.

What it looks like in practice

The flagship, aida, run on a single task:

It never picked your architecture for you. It just refused to skip the scope, the existing-solution check, the tests, or the review, and it left everything it did on disk, so a decision that later looks wrong is a file you open. That is the idea behind all of these plugins.

None of it is perfect. The AI still slips a step past us now and then, and when it does, that gap becomes the next gate we add. See PHILOSOPHY.md for the fuller picture.

Installation

Each plugin's README has its own requirements, worked examples, and a docs/usage.md. Versions and component counts live there and in each CHANGELOG.md, not here, so this page does not go stale.

Using these plugins outside Claude Code

The skills conform to the open agentskills.io standard and work in Cursor, Codex CLI, VS Code Copilot, Gemini CLI, Cline, OpenCode, and more. Commands, agents, and hooks are Claude-Code-specific by format but can be emulated. See PORTABILITY.md for the full guide and CURSOR.md for the highest-fidelity option.

The plugins, by the problem each solves

Building software

aida: *The AI jumps straight to code: it skips understanding the problem, misses a library that already exists, drifts from your standards, and forgets the tests.* Runs any coding task through a contract, then research, design, implementation against frozen tests, review, and completion. The checks are scripts…