ai-skills
CLAUDE.md1 starsChanged 3 months agoMITView on GitHub
Global preferences (identity, communication style, tools, standards, workflow) are defined in ~/.claude/CLAUDE.md and apply here. This file adds only repo-specific instructions on top of those. A curated collection of AI skill prompt files for use with Claude Code (and other AI assistants). Skills are structured prompt templates that Claude Code loads on demand via the Skill tool to provide specialized, repeatable behaviors. This repo is compatible with skillshare, a CLI tool for managing and syncing skills across AI agents. The…
- Commits and pushes
In this repository16 other files it gives its agents.
- AGENTS.mdAGENTS.md
- anger-translatorSkill
- cache-moneySkill
- codebase-memoirSkill
- codex-reviewSkill
- goalpostSkill
- grand-tourSkill
- identity-theftSkill
- relocate-sessionSkill
- right-stuffSkill
- rubicon-wordpress-version-labSkill
- second-lookSkill
- secret-santa-generatorSkill
- showdownSkill
- star-chartSkill
- work-evidence-researchSkill
The file
Show its rules and text
We split this CLAUDE.md into the rules it gives an agent
16 rules. About 0 are common, 4 are distinctive, and 12 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
12
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 61
Write skill content in plain Markdown.
- Distinctive: line 62
Keep skills focused on a single task or workflow.
- Very specific: line 63
Every skill directory also contains a README.md (thin, human-facing: a one-line description, the skillshare install command, and links to …
- Distinctive: line 81
Create or reference an issue.
- Very specific: line 83
Create skills/<skill-name>/SKILL.md with valid YAML frontmatter (name + description required.
- Very specific: line 84
Add skills/<skill-name>/README.md (thin stub) and skills/<skill-name>/CHANGELOG.md (initial version entry, no ## [Unreleased] section.
- Very specific: line 85
Put the skill under release-please.
- Very specific: line 86
Update the skills table in root README.md.
- Very specific: line 87
Commit with feat: prefix; open a PR on GitHub.
- Very specific: line 109
Create plugins/<plugin-name>/.claude-plugin/plugin.json (name required.
- Very specific: line 110
Add plugins/<plugin-name>/README.md and plugins/<plugin-name>/CHANGELOG.md (Keep a Changelog.
- Very specific: line 112
Put the plugin under release-please.
- Very specific: line 113
Leave the root CHANGELOG.md alone.
- Very specific: line 114
Run bash scripts/validate-skills.sh.
- Very specific: line 126
Update CHANGELOG.md before tagging a release.
- Distinctive: line 127
Create GitHub releases only when public skill content changes.
≈ “Common” is an estimate: at least 1 in 5 of the 28,029 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Global preferences (identity, communication style, tools, standards, workflow) are defined in ~/.claude/CLAUDE.md and apply here. This file adds only repo-specific instructions on top of those.
Repository Purpose
A curated collection of AI skill prompt files for use with Claude Code (and other AI assistants). Skills are structured prompt templates that Claude Code loads on demand via the Skill tool to provide specialized, repeatable behaviors.
This repo is compatible with skillshare, a CLI tool for managing and syncing skills across AI agents.
The canonical repository is public on GitHub at github.com/rubicon/ai-skills; a private archive of the pre-migration history is retained on Forgejo. Everything in this repository is public. Skills that cannot be published live in the separate private repository ai-skills-private, which is its own repo and not part of this one.
Structure
skills/
├── secret-santa-generator/
│ ├── SKILL.md
│ ├── README.md
│ └── CHANGELOG.md
└── <skill-name>/
├── SKILL.md # required — the skill definition (model-facing)
├── README.md # required — human-facing overview + install
├── CHANGELOG.md # required — per-skill version history
└── references/ # optional — supporting files
plugins/
└── <plugin-name>/ # Claude Code plugin (self-contained)
├── .claude-plugin/plugin.json # required — manifest (SemVer version)
├── README.md # required
├── CHANGELOG.md # required — Keep a Changelog
└── skills/ commands/ agents/ scripts/ # optional plugin components
.claude-plugin/
└── marketplace.json # plugin marketplace manifest (name: rubicon)
README.md
CHANGELOG.md
CLAUDE.mdEach skill is a directory inside skills/ containing a SKILL.md file, plus a README.md and CHANGELOG.md (see Skill File Conventions). This structure satisfies both Claude Code's Skill tool and the skillshare CLI. Flat .md files in skills/ are not recognized by skillshare and should not be used.
Claude Code plugins live in their own independent tree under plugins/, listed in the repo-root .claude-plugin/marketplace.json. Plugins are self-contained (no ../ references), so a plugin's bundled skills/agents/scripts live inside the plugin, not in the top-level skills/. See the Plugin Structure Rule in docs/process/ai-skills-repo-overlay.md.
Skill File Conventions
- One skill per directory; directory name must be kebab-case (e.g.,
secret-santa-generator/). - The skill definition lives at
skills/<skill-name>/SKILL.md— skillshare requires exactly this filename. - Each
SKILL.mdmust begin with YAML frontmatter including at minimumnameanddescription:
---
name: skill-name
description: >-
One-line (or multi-line) description used by Claude to decide when to
invoke this skill. Include trigger phrases here.
---- Optional frontmatter fields:
author,version,targets(e.g.,[claude, cursor]). Theversionfield uses full SemVer (MAJOR.MINOR.PATCH, e.g.1.0.0). - Write skill content in plain Markdown. Use headers, bullet lists, and code blocks as needed.
- Keep skills focused on a single task or workflow. Split if a skill grows unwieldy.
- Every skill directory also contains a
README.md(thin, human-facing: a one-line description, the skillshare install command, and links toSKILL.mdandCHANGELOG.md— do not duplicateSKILL.md's behavioral content) and aCHANGELOG.md(Keep a Changelog form; its latest version heading matches theversion:inSKILL.md).
Documentation & Changelogs
Root README.md — when adding a new skill, add a row to the skills table:
| [skill-name](skills/skill-name/SKILL.md) | One-line description |Per-skill README.md — a thin, human-facing stub: a one-line description, the skillshare install command, and links to SKILL.md and CHANGELOG.md. Keep behavior in SKILL.md; do not copy it into the README.
Per-skill CHANGELOG.md — the detailed version history for that skill, in Keep a Changelog form. Its latest version heading matches the version: field in the skill's SKILL.md.
Root CHANGELOG.md — generated by release-please from Conventional Commits PR titles (release-please-config.json, package .). Do not hand-edit it, and do not add a line for a new skill or plugin. The per-skill and per-plugin CHANGELOG.md is where the hand-written first entry goes.
Adding a Skill
- Create or reference an issue; note the issue number.
- Create branch:
dev/<issue-number>-<short-kebab-description> - Create
skills/<skill-name>/SKILL.mdwith valid YAML frontmatter (name+descriptionrequired;versionin SemVer). - Add
skills/<skill-name>/README.md(thin stub) andskills/<skill-name>/CHANGELOG.md(initial version entry, no## [Unreleased]section; see Versioning and Changelogs indocs/process/ai-skills-repo-overlay.md). - Put the skill under release-please. In
release-please-config.jsonadd"skills/<skill-name>": { "package-name": "<skill-name>", "extra-files": ["SKILL.md"] }; in.release-please-manifest.jsonadd"skills/<skill-name>": "<starting version>"; and mark theSKILL.mdversion lineversion: 0.1.0 # x-release-please-version.scripts/validate-skills.shfails when either entry is missing. - Update the skills table in root
README.md. Leave the rootCHANGELOG.mdalone; release-please generates it. - Commit with
feat:prefix; open a PR on GitHub.
Pre-merge checklist:
- Skill lives at
skills/<skill-name>/SKILL.md(directory, not flat file) - YAML frontmatter is valid and includes
nameanddescription;versionis SemVer (MAJOR.MINOR.PATCH) - Skill directory has
README.mdandCHANGELOG.md - Skill has a
release-please-config.jsonentry, a.release-please-manifest.jsonentry, and the# x-release-please-versionmarker on itsversion:line - Root
README.mdskills table updated; rootCHANGELOG.mduntouched (release-please owns it) - No personal data in installed content —
bash scripts/check-no-personal-data.shpasses
Examples in skills/ and plugins/ are invented, never observed. Everything there is published
and installed on other people's machines. A directory tree, file count, document title, or path
copied from a real folder publishes that folder's layout, and it reads as an ordinary example, so
nothing flags it later. check-no-personal-data.sh catches absolute home paths and real email
addresses; it cannot catch content that is personal by virtue of being accurate. Write the example,
do not transcribe it.
Adding a Plugin
- Create or reference an issue; note the issue number.
- Create branch:
dev/<issue-number>-<short-kebab-description> - Create
plugins/<plugin-name>/.claude-plugin/plugin.json(namerequired;versionin SemVer when present). - Add
plugins/<plugin-name>/README.mdandplugins/<plugin-name>/CHANGELOG.md(Keep a Changelog; no## [Unreleased]section, see Versioning and Changelogs indocs/process/ai-skills-repo-overlay.md). Bundled skills/commands/agents/scripts live inside the plugin (self-contained — no../). - Append the plugin to
.claude-plugin/marketplace.json:{ "name": "<plugin-name>", "source": "./plugins/<plugin-name>" }. - Put the plugin under release-please. In
release-please-config.jsonadd"plugins/<plugin-name>": { "package-name": "<plugin-name>", "extra-files": [{ "type": "json", "path": ".claude-plugin/plugin.json", "jsonpath": "$.version" }] }; in.release-please-manifest.jsonadd"plugins/<plugin-name>": "<starting version>".scripts/validate-skills.shfails when either entry is missing. - Leave the root
CHANGELOG.mdalone; release-please generates it. - Run
bash scripts/validate-skills.sh; commit withfeat:prefix; open a PR on GitHub.
See the Plugin Structure Rule in docs/process/ai-skills-repo-overlay.md for the full convention.
Process Policy
This repo follows the general repository process policy. Repo-specific rules are in docs/process/ai-skills-repo-overlay.md.
Key points:
- Issue →
dev/<issue-number>-<short-kebab-description>branch → PR → merge. No direct pushes tomain. - Semantic commit prefixes:
feat:,fix:,docs:,chore:. - Update
CHANGELOG.mdbefore tagging a release. - Create GitHub releases only when public skill content changes. Releases should reflect skill additions, updates, or removals that affect end users.
Git Hosting
Canonical host is GitHub — origin is git@github.com:rubicon/ai-skills.git (SSH; commit signing via the 1Password SSH agent). main is protected: PRs only, required CI checks, signed commits, linear history, no force-push. Ask before committing or pushing.
Issues and PRs use the GitHub gh CLI:
gh issue create -t "<plain descriptive title>" -b "<body>"
gh pr create --base main --head dev/<n>-<slug> -t "<conventional-commits title>" -b "<body>"A private archive of the pre-migration history lives on Forgejo (git.daxdavis.com); it is a backup only, not the working host. Anything local you do not want published goes in CLAUDE.local.md (git-ignored) — do not put personal or machine-specific notes in this committed file.
Migration record: docs/runbooks/github-migration.md.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

