Back to search
obie-skills
SkillGitHubIncomplete
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/obie--skills)HTML
<a href="/listings/obie--skills"><img src="/api/listings/obie--skills/badge.svg" alt="findsafeskills safety rating" /></a>Install
i
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
Add this marketplace to your Claude Code:
/plugin marketplace add obie/skills
Then install individual skills:
/plugin install better-stimulus@obie-skills
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 183 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
- {
- "name": "obie-skills",
- "owner": {
- "name": "Obie Fernandez",
- "email": "obiefernandez@gmail.com"
- },
- "metadata": {
- "description": "Curated collection of production-ready Claude Code skills",
- "version": "1.0.0"
- },
- "plugins": [
- {
- "name": "better-stimulus",
- "source": "./skills/better-stimulus",
- "description": "Apply Better Stimulus best practices for writing maintainable, reusable StimulusJS controllers following SOLID principles",
- "version": "1.0.0",
- "author": {
- "name": "Obie Fernandez"
- },
- "keywords": [
- "stimulus",
- "hotwire",
- "rails",
- "javascript",
- "best-practices",
- "solid"
- ],
- "category": "development",
- "homepage": "https://github.com/obie/skills",
- "license": "MIT"
- },
- {
- "name": "rails-activity-timeline",
- "source": "./skills/rails-activity-timeline",
- "description": "Add polymorphic activity timelines with live Turbo Stream updates to any Rails model",
- "version": "1.0.0",
- "author": {
- "name": "Obie Fernandez"
- },
- "keywords": [
- "rails",
- "hotwire",
- "turbo-streams",
- "activity-log",
- "audit-trail",
- "timeline",
- "action-cable"
- ],
- "category": "development",
- "homepage": "https://github.com/obie/skills",
- "license": "MIT"
- },
- {
- "name": "rails-tiptap-autosave",
- "source": "./skills/rails-tiptap-autosave",
- "description": "Add Tiptap rich text editing with debounced autosave to Rails models using Stimulus",
- "version": "1.0.0",
- "author": {
- "name": "Obie Fernandez"
- },
- "keywords": [
- "rails",
- "tiptap",
- "stimulus",
- "rich-text",
- "autosave",
- "markdown",
- "hotwire"
- ],
- "category": "development",
- "homepage": "https://github.com/obie/skills",
- "license": "MIT"
- },
- {
- "name": "deletion-test",
- "source": "./skills/deletion-test",
- "description": "Run the regenerative-software deletion test on a component: extract intent and decision records, write a durable contract spec, mutation-test both oracles, regenerate the implementation in isolated arms, score by behavioral diff, and report evaluation gaps",
- "version": "0.1.0",
- "author": {
- "name": "Obie Fernandez"
- },
- "keywords": [
- "regenerative-software",
- "deletion-test",
- "mutation-testing",
- "mutant",
- "provenance",
- "decision-records",
- "evaluations",
- "ruby",
- "rails",
- "feelings"
- ],
- "category": "development",
- "homepage": "https://github.com/obie/skills",
- "license": "MIT"
- },
- {
- "name": "graphs-and-shit",
- "source": "./skills/graphs-and-shit",
- "description": "Pull the latest data from a survey, sheet, API, database or file and turn it into a shareable, editable Claude Doc with hand-drawn charts, each one screenshot-checked; refreshes the doc when the data changes",
- "version": "0.1.0",
- "author": {
- "name": "Obie Fernandez"
- },
- "keywords": [
- "charts",
- "data-visualization",
- "reports",
- "claude-docs",
- "artifacts",
- "surveys",
- "google-forms"
- ],
- "category": "productivity",
- "homepage": "https://github.com/obie/skills",
- "license": "MIT"
- },
- {
- "name": "beat-the-benchmark-website",
- "source": "./skills/beat-the-benchmark-website",
- "description": "Build a website that measurably beats a named benchmark site: audit the benchmark, owner voting galleries, a parallel agent build with one integrator, QC and Codex review of an immutable preview, then a scope freeze and byte-parity production ship",
- "category": "development",
- "keywords": [
- "website",
- "qc",
- "playwright",
- "cloudflare-pages",
- "agent-teams",
- "review",
- "astro"
- ],
- "version": "1.0.0",
- "author": {
- "name": "Obie Fernandez"
- },
- "homepage": "https://github.com/obie/skills",
- "license": "MIT"
- },
- {
- "name": "ai-character-imaging",
- "source": "./skills/ai-character-imaging",
- "description": "Keep one AI character or persona consistent across stills and motion loops: canon identity pack, prompt slots, owner review rounds, filter workarounds, waxy-free upscaling, feathered patch merges and seamless Kling loops",
- "category": "creative",
- "keywords": [
- "ai-images",
- "character-consistency",
- "gpt-image",
- "kling",
- "fal",
- "persona",
- "video-loops"
- ],
- "version": "1.0.0",
- "author": {
- "name": "Obie Fernandez"
- },
- "homepage": "https://github.com/obie/skills",
- "license": "MIT"
- },
- {
- "name": "decision-docket",
- "source": "./skills/decision-docket",
- "description": "Keep one standing page of every decision you need from the user, each answerable cold with its context, options, costs and gains, your recommendation and what happens if they do nothing; they record answers on the page, you act on them, log the ruling and remove the card",
- "category": "productivity",
- "keywords": [
- "decisions",
- "questions",
- "artifacts",
- "async",
- "autonomous-agents",
- "human-in-the-loop",
- "docket"
- ],
- "version": "1.0.0",
- "author": {
- "name": "Obie Fernandez"
- },
- "homepage": "https://github.com/obie/skills",
- "license": "MIT"
- }
- ]
- }
README.md · README · 169 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
- # Claude Code Skills by Obie
- Production-ready skills for enhanced Claude Code development workflows.
- ## Installation
- Add this marketplace to your Claude Code:
- ```shell
- /plugin marketplace add obie/skills
- ```
- Then install individual skills:
- ```shell
- /plugin install better-stimulus@obie-skills
- ```
- ## Available Skills
- ### Better Stimulus
- Apply opinionated best practices from [betterstimulus.com](https://betterstimulus.com/) when writing or refactoring Stimulus controllers.
- **When to use:**
- - Writing new Stimulus controllers
- - Refactoring existing Stimulus code
- - Reviewing Stimulus controller architecture
- - Debugging inter-controller communication
- **Key patterns:**
- - Values API for reactive state management
- - Static classes for configurable CSS
- - Single Responsibility Principle
- - Minimal connect() usage
- - Declarative event registration
- **Invoke with:** `/better-stimulus` or Claude invokes automatically when working with Stimulus code
- ### MCP OAuth Setup
- Implement zero-configuration OAuth for MCP (Model Context Protocol) server connections using Dynamic Client Registration (RFC 7591) and Authorization Server Metadata Discovery (RFC 8414). Battle-tested with Linear, Sentry, and Granola.
- **When to use:**
- - Building admin UIs for managing MCP server connections with OAuth
- - Integrating with third-party MCP providers
- - Implementing MCP Streamable HTTP transport with authenticated tool sync
- - Adding OAuth to an existing MCP connector management system
- **What it covers:**
- - RFC 8414 metadata discovery + RFC 7591 dynamic client registration
- - PKCE (S256) authorization flow with signed state
- - Token exchange, storage, and refresh (shared + per-agent modes)
- - MCP Streamable HTTP tool sync (session handshake, SSE parsing)
- - UI patterns (Stimulus forms, Connect vs Sync logic)
- - Common failure modes and fixes (Turbo Drive, route helpers, SSE)
- **Invoke with:** `/mcp-oauth-setup` or Claude invokes automatically when implementing MCP OAuth
- ### Deletion Test
- Run the deletion test from Chad Fowler's *Regenerative Software* on one component: could we delete this and rebuild it from what the system holds outside the code? Extracts intent and decision records from code, history, and tickets; writes a durable contract spec that asserts promises rather than pinned numbers; mutation-tests both oracles; deletes the implementation and regenerates it in isolated arms from the intent alone; scores every candidate by behavioral diff; reports what the code knew that nothing else did.
- **When to use:**
- - Making a component safely replaceable, or finding out whether it already is
- - Extracting the reasons buried in a legacy module's comments, commits, and tickets
- - Checking whether a test suite is an oracle or a museum of the current implementation
- - Establishing a baseline before adopting regeneration tooling
- **What it covers:**
- - Intent document and decision-record formats, with the sparse variant for a decisions-withheld arm
- - Durable contract specs: properties over grids, the language-change membership test, degenerate inputs
- - mutant two-pass scoring with a survivor classification table
- - Isolated regeneration arms in git worktrees with a regenerator brief
- - Behavioral-diff scoring scripts and a findings template
- - Optional survivor triage with a calibrated decision model (feelings / Jev)
- **Invoke with:** `/deletion-test` or Claude invokes automatically when asked to regenerate a module from spec or make a component replaceable
- ### Graphs and Shit
- Pull the latest data from wherever it lives and turn it into a shareable, editable Claude Doc with hand-drawn charts. The lead states the answer with its numbers. Each section gets one chart for one finding, and every chart is checked in a screenshot before the doc is called done. Run it again later and it refreshes the same doc with new numbers.
- **When to use:**
- - Survey results that need to go in front of a team
- - A spreadsheet or query result that people will discuss and comment on
- - Numbers that keep changing, such as responses trickling in, where the doc should be refreshed rather than rebuilt
- **What it covers:**
- - Pulling fresh data (Google Forms, Sheets, CSV, a database, an API) and tallying it with a script
- - A short brief per chart: the reader, the finding, the form a pro would draw
- - Creating the doc from the Docs Artifact type and landing an outline the reader can watch fill
- - Hand-rolled `claude.Visualize` charts: grey first, one accent, values on the marks, everything computed from rows
- - Screenshot review and redraws
- - Keeping anonymous data anonymous
- - Refreshing chart rows and prose numbers without clobbering people's edits
- Requires the Artifact tool and the Claude Docs connector. Without the connector it falls back to a single HTML Artifact.
- **Invoke with:** `/graphs-and-shit <data source>` or Claude invokes automatically when asked to chart results into something shareable
- ### beat-the-benchmark-website
- Build a site that beats a named benchmark site on every dimension the owner cares about, and prove it.
- **When to use:** a new or rebuilt website with a reference site to beat, and an owner who wants evidence, not opinions.
- **Key patterns:**
- - Audit the benchmark into a scored QC matrix before building
- - One generated source of truth; owner voting galleries for every visual choice
- - Parallel builders, one integrator; QC and Codex review an immutable preview
- - Scope freeze, then byte-parity check of production against the verified build
- - Playwright scripts: screenshot matrix, decoded image ratio, tap targets, forbidden-token sweep, JSON-LD
- **Invoke with:** `/beat-the-benchmark-website <site> vs <benchmark>`
- ### ai-character-imaging
- Keep one synthetic character recognisably the same across many images and short video loops.
- **When to use:** a virtual artist, mascot or brand persona that needs a consistent face across plates and motion.
- **Key patterns:**
- - Canon identity pack of the newest approved face, plus an identity paragraph
- - Slot prompts; refs carry identity only; GPT Image edit with refs
- - Owner review rounds; promote approved picks to ref #1
- - Upscale with face enhancement off; patch defects with a feathered merge
- - Kling image-to-video loops with the seam scored by frame difference
- **Invoke with:** `/ai-character-imaging` or Claude invokes automatically for consistent-character generation
- ### Decision Docket
- Keep one standing page of every decision Claude needs from you, instead of questions buried in chat replies. Each card can be answered cold: what happened, why it's your call, the options with their cost and gain, Claude's recommendation, and what happens if you do nothing. You pick an option and record it on the page; on its next check-in Claude reads the answer, acts on it, logs the ruling and removes the card, so the page only ever shows what is still waiting on you.
- **When to use:**
- - Long-running or autonomous work where Claude keeps going while you're elsewhere
- - You move between many threads and lose questions asked inline
- - Sorting a list of items (keep, hide, cut) or auditing calls Claude made without you
- **What it covers:**
- - A ready-made docket page (an Artifact with a small database): a sticky index, one card per decision, light and dark themes, a phone layout, drafts kept while you type
- - The question format and how to write a card someone can decide from with no other context
- - The check-in loop: read answers, act, log the ruling, delete the card pinned to its version
- - FYI "calls" cards, item lists with their own choices, owner-only or team write access
- **Invoke with:** `/decision-docket` or Claude invokes automatically when it has a decision to put to you during long-running work
- ## How It Works
- Skills are stored in `skills/<skill-name>/SKILL.md` format. Each skill:
- - Contains YAML frontmatter describing when to use it
- - Includes comprehensive instructions for Claude
- - May reference supporting documentation files
- - Can be invoked manually with `/skill-name` or automatically by Claude
- ## Contributing
- Want to add your own skills to this marketplace?
- 1. Fork this repository
- 2. Create a new skill directory under `skills/`
- 3. Write your `SKILL.md` following the [Agent Skills standard](https://agentskills.io)
- 4. Add an entry to `.claude-plugin/marketplace.json`
- 5. Submit a pull request
- ## License
- MIT - See LICENSE file for details
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.