luckycold/agent-skills

Luke-authored portable skills for infrastructure, automation, authentication, and agent workflow maintenance. This is a **public** repository: procedures use placeholders, while private operational values remain in an ignored local ~/.agents/private-context.md.
First-class consumers include Labby (Cursor home-lab bot), Codex, Cursor, Claude Code, OpenCode, OpenClaw, and Hermes (historical Home Assistant add-on agent). Prefer portable skills; Hermes-specific HA-addon paths stay in infrastructure-hygiene references.
Canonical checkout
Clone this repository as the shared cross-agent skills area:
Install or refresh with the Agent Skills CLI (Codex, Claude Code, Cursor, OpenCode):
Hermes (historical HA add-on) can also load the checkout as an external directory:
skills.external_dirs makes the checked-out packages available to Hermes and writable in place. The tap adds the GitHub repository as a discovery/install source; it does not replace the writable checkout.
Skills
- direct-action-preferences
- infrastructure-hygiene
- minecraft-steam-gamescope
- native-desktop-temporary
- ntfy-ops
- personal-skill-maintenance
- proton-pass-cli
- proxmox-homelab
- steam-frame-linux-networking
- steam-hyprland-scaling
- tasker-automation
- temporary-support-watcher-cleanup
- truenas-custom-apps
Safety and validation
Before committing:
The repository's safety check rejects common secret formats, private keys, private network addresses, private-domain markers, emails, and tracked private-context files. It also runs scripts/check_skill_size.py, which fails any skills/*/SKILL.md over 200 lines and warns above 100. GitHub Actions runs the same check on pushes and pull requests.