A small collection of agent skills I use day-to-day. Harness-agnostic: each skill is just a folder with a SKILL.md (and optional assets), portable across any agent runtime that supports the SKILL convention — Claude Code, Codex, Cursor, etc.
A directory containing a SKILL.md file with YAML frontmatter (name, description) and a body the agent loads as instructions when the skill is invoked. Optional sibling files (scripts, templates, references) live alongside.
<skill-name>/
SKILL.md # frontmatter + agent instructions
assets/ # optional: scripts, templates, etc.
| Skill | What it does |
|---|---|
handoff |
Durable project state across sessions — state machine over GitHub issues (preferred) or repo docs, with checkpoints, drift checks, and a feature-status vocabulary. |
md |
Render markdown to a styled HTML page in the user's browser, using GitHub's /markdown API for the real GFM pipeline (tables, task lists, syntax highlighting). |
How you install depends on your agent harness:
- Claude Code: drop the skill folder into
~/.claude/skills/<name>/(user-scope) or<project>/.claude/skills/<name>/(project-scope). - Codex / others: consult your runtime's docs for the skills directory.
A common pattern is to clone this repo and symlink individual skills into the runtime's skills directory, so updates from git pull propagate.
git clone https://github.com/danseely/skills.git ~/src/skills
ln -s ~/src/skills/md ~/.claude/skills/md
ln -s ~/src/skills/handoff ~/.claude/skills/handoff