Reusable LLM agent skills for Codex, Claude Code, and similar workflows.
This repository is a folder-based collection of installable skills. Each skill lives in its own top-level directory:
handover/- future skills as sibling folders
This repo is for:
- developers building their own agent workflows
- people who want reusable skills across multiple repos
- anyone using Codex or Claude Code with custom skills
Each skill is a self-contained folder:
README.md
handover/
another-skill/
Typical skill contents:
skill-name/
├── SKILL.md
└── templates/
For now, installation is copy-based.
Copy a skill folder into ~/.codex/skills/.
Example:
cp -R handover ~/.codex/skills/handoverThen restart Codex.
Copy a skill folder into ~/.claude/skills/.
Example:
cp -R handover ~/.claude/skills/handoverThen restart Claude Code.
Create a precise continuation packet when you want to stop work and pass the task to another LLM.
What it does:
- writes one timestamped Markdown handoff file
- embeds a compact JSON block in that file
- captures current status, completed work, remaining work, verification, and next steps
- preserves user-supplied handoff instructions as:
- required instructions
- optional notes
Use it when you explicitly want to hand a task from one agent to another.
These skills are meant to be:
- practical
- portable
- easy to read
- easy to copy into your own setup
More skills will be added over time as the library grows.