Skill Codex
Design and run evaluations for prompts, models, retrieval, routing, tool use, structured outputs, and multi-step agent trajectories. Use when AI behavior changes or when failures should become regression datasets.
Skill Codex
Author executable Given-When-Then behavior contracts for observable product, API, CLI, data, ML, or agent behavior. Use when acceptance criteria need examples or when behavior changes should be described before implementation.
Skill Codex
Initialize a cloned template repository for a concrete project. Use when the repository still contains template placeholders, the project purpose or stack is unclear, the user asks to fill AGENTS.md/README/BOARD/docs, or a new project is being started from this template.
Skill Codex
Diagnose and fix defects with a permanent reproducer. Use for bug reports, broken behavior, incorrect outputs, unsafe tool behavior, failed integrations, data errors, prompt failures, or any issue that must not recur.
Skill Codex
Enforce test-first development for deterministic logic, behavior changes, refactors, and implementation tasks. Use when adding or changing code behavior, fixing a defect, or implementing a board task that needs Red, verified failure, minimal Green, verified success, and safe refactoring.
Skill Codex
A workflow for creating a ZIP archive containing only Git-untracked files. Untracked files are new files Git is not yet tracking.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: