Vibes

skills

Jon Hill's public collection of portable Agent Skills — self-contained, model- and harness-agnostic instructions an AI coding agent loads on demand. Every skill here is individually installable; nothing in this repository is specific to any one harness, and nothing here depends on private tooling or evidence.

Install

Browse the collection:

npx skills add jonhill90/skills --list

Install one or more specific skills into the current project:

npx skills add jonhill90/skills --skill tmux --skill github-cli

npx skills pins installs by content hash in skills-lock.json, so a project's skill set stays reproducible. See the skills CLI for the full command reference.

As an Agent Plugin

This repository is also an Agent Plugins 1.0.0 plugin: plugin.json at the root, 24 skills at skills/<name>/SKILL.md, which is the standard's own discovery convention. Any conformant client can consume the collection whole, with no bespoke tooling.

That is a portability claim, not a local one. Claude Code reads its own manifest at .claude-plugin/plugin.json and does not look for this file, so adding it changed nothing for Claude Code users today. npx skills above remains the way to install individual skills, and nothing here replaces it.

Skills in this collection

Skill	Purpose
close-the-loop	Confirm you have everything needed to finish a change before starting it
create-skill	Design, create, and validate portable Agent Skills
dispatching-subagents	Decide whether to delegate to subagents and verify their output with external evidence
failing-test-first	Reproduce a bug with a failing test before fixing it
github-cli	Manage GitHub PRs, issues, workflows, actions, and releases via gh
linear	Manage Linear issues, teams, and projects via the Linear CLI
memory-conventions	Read and write durable agent memory in a personal Obsidian vault
obsidian	Read, write, search, and manage notes in Obsidian vaults
primer	Orient in an unfamiliar codebase before starting work
safe-deletion	Verify contents match their described purpose before deleting anything
sanity-check	Build a second-opinion reviewer prompt for high-cost reasoning
supervised-lane-loop	Run a long-lived supervisor loop over one or more worker-agent lanes
tmux	Operate tmux safely from an agent: pane targeting, verified input, recovery

Each skill's description frontmatter is the actual trigger contract — read the skill itself for exact wording and preconditions.

Where a skill belongs

Most skills should not live in this repository. Decide placement first:

Situation	Where it goes
Useful across many unrelated projects, every day	a public collection like this one
Only true in one repository	that repo's own .claude/skills/ or .agents/skills/
Needed once, or maintained by someone else	nothing installed — npx skills use <package>@<skill>

Authoring contract

Each skill lives at skills/<name>/SKILL.md:

skills/example-skill/
├── SKILL.md
├── scripts/       Optional deterministic, tested helpers
├── references/    Optional detail loaded on demand
└── assets/        Optional output resources

Portable frontmatter is name and description (plus optional license, compatibility, metadata, allowed-tools). The directory name must match name. Keep SKILL.md under 500 lines and move detail into directly linked references/. Full conventions: AGENTS.md.

Validate

python3 scripts/validate_repository.py
python3 -m unittest discover -s tests -v

CI runs both on every pull request and on pushes to main.

Content boundaries

• This repository holds only portable skill content and the minimal validation/tests/CI needed to keep it correct.
• Personal harness configuration — canonical instructions, hooks, agents, settings, MCP declarations, install/sync tooling — lives in a separate personal harness repository that consumes this collection; it is not vendored here.
• Behavioral evaluation methodology, scenarios, transcripts, and results are private and are not published in this repository. Where a skill references past evidence, it states what happened and when without a link to private material.
• Employer-owned or project-specific material is never copied here.

See AGENTS.md for contribution rules.