Vibes
Transforms Claude Desktop into a conversational development environment using distributed MCP servers.
Install
Install and configure the MCP from https://github.com/jonhill90/vibes now. Follow the repository's installation instructions, ask me for anything you can't complete yourself, and verify its tools load.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, every skill at skills/<name>/SKILL.md (the generated table below is the
count), which is the standard's own discovery convention.
Any conformant client can consume the collection whole, with no bespoke
tooling.
As a Claude Code plugin
.claude-plugin/ makes this repository its own Claude Code marketplace with
one plugin, jonhill90-skills, that contains every skill:
claude plugin marketplace add jonhill90/skills
claude plugin install jonhill90-skills@jonhill90
The plugin has no pinned version, so each install and update takes the
latest commit on main. It loads the whole collection, which adds roughly
5.6k tokens of skill descriptions to every session (measured 2026-09-30 with
claude plugin details). To take only a few skills, use npx skills above.
Skills in this collection
Generated from 42 skills; do not hand-edit.
Regenerate with python3 scripts/skills_table.py; verify with --check.
Upload audit PASS = no local-machine or CLI assumptions detected; FAIL = review before using without a shell.
| Skill | Upload audit |
|---|---|
adopt-or-build | FAIL |
ask-a-council | PASS |
close-the-loop | PASS |
create-skill | FAIL |
decide-by-variant | PASS |
derive-independently-then-compare | PASS |
determine-intent | PASS |
determine-signals | PASS |
devils-advocate | PASS |
dispatch-brief | PASS |
dispatching-subagents | PASS |
distill | PASS |
durable-fact-before-label | PASS |
failing-test-first | PASS |
github-cli | FAIL |
keep-me-honest | PASS |
linear | FAIL |
loop-contract | PASS |
loop-memory | FAIL |
mechanize | PASS |
memory-conventions | FAIL |
mine-transcripts | FAIL |
notify | FAIL |
obsidian | FAIL |
organize-repository | PASS |
plan-parallel-execution | PASS |
prd | PASS |
primer | FAIL |
progressive-disclosure | PASS |
prompt-corpus | FAIL |
refuse-invented-identity | PASS |
research-the-limit | PASS |
safe-deletion | PASS |
sanity-check | PASS |
spec | PASS |
spec-driven-development | PASS |
supervised-lane-loop | FAIL |
tdd | PASS |
test-in-the-consumer-context | PASS |
tmux | FAIL |
verify-the-instrument | PASS |
wire-it-when-you-write-it | PASS |
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.
- There is no behavioral eval system here for now; a community-maintained replacement is tracked in #309. 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.