Skip to content

Course-Studio/rare-skills

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

42 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Rare Skills 💎

Battle-tested AI skills for daily developer workflows.

Installation

Add the marketplace and install the plugin:

/plugin marketplace add Course-Studio/rare-skills
/plugin install rd@rare-skills

After installation, all 12 skills are available in any project. Run /plugin and check the Installed tab to verify.

Updating

/plugin marketplace update rare-skills
/plugin update rd@rare-skills

Uninstalling

/plugin uninstall rd@rare-skills
/plugin marketplace remove rare-skills

Skills

Skill Description
/address-pr-feedback Process and address code review feedback with technical rigor
/brainstorm Explore design options through structured discussion and produce a saved spec
/browse Automate browser interactions, navigate websites, fill forms, take screenshots, and manage sessions via Playwright
/commit Create clean, atomic commits following project conventions
/create-unit-tests Generate unit tests following company testing standards (Vitest/RTL for frontend, pytest/factory-boy for backend)
/generate-qa-steps Generate QA testing steps from code changes and spec files
/open-pr Create or update pull requests with rationale-focused descriptions
/reflect Retrospective review of the current session to surface mistakes, friction, and improvements
/skill-creator Create new skills, run evals, benchmark performance, and iteratively improve skill quality
/test-in-browser Test UI in the browser, verify design and layout, inspect console, and check visual appearance via Playwright
/ticket Create, read, update, and comment on Linear tickets
/timesheet Reconstruct and record weekly timesheets from Google Calendar, Slack, Granola, and Gmail

/address-pr-feedback

Process and address code review feedback with technical rigor.

  • Reads all PR comments via GitHub CLI
  • Groups by severity: must-fix, suggestion, question, nit
  • Evaluates suggestions against design docs when available
  • Confirms evaluation with developer before implementing changes

/brainstorm

Explore design options through structured discussion and produce a saved spec.

  • Explores project context before asking questions
  • Asks one clarifying question at a time, multiple choice preferred
  • Proposes 2-3 approaches with trade-off comparison table and recommendation
  • Builds spec section by section with approval after each
  • Saves spec as local file or Linear ticket comment
  • Does NOT trigger implementation -- the spec is the deliverable

/browse

Automate browser interactions using the Playwright CLI.

  • Complete CLI command reference (navigation, clicks, fills, screenshots, tabs, storage, network, devtools)
  • Persistent sessions preserve login state across runs
  • Headed mode so the developer can see browser activity
  • Session management for concurrent browser instances
  • 7 reference docs covering session management, test generation, video recording, tracing, request mocking, storage state, and running code

/commit

Create clean, atomic commits that follow the project's conventions.

  • Reviews staged and unstaged changes
  • Proposes atomic commits in conventional format
  • Blocks secrets from being committed (.env, API keys, tokens, private keys)
  • Confirms before executing
  • No AI attribution in commit messages

/create-unit-tests

Generate unit tests that cover business-critical logic following the project's testing standards.

  • Auto-detects stack: Vitest + React Testing Library for frontend, pytest + factory-boy for backend
  • Prioritizes revenue-impacting paths, auth/security, and data transformations
  • Follows Arrange/Act/Assert pattern with descriptive test names
  • Covers CMS nullable field handling, webhook testing patterns, and mocking best practices
  • Suggests integration tests when code involves API routes or DB-backed services

/generate-qa-steps

Generate structured QA testing steps from code changes and spec files.

  • Works standalone or chained from /open-pr
  • Accepts PR URL as argument: /generate-qa-steps https://github.com/org/repo/pull/123
  • Automatically picks up brainstorm spec files from docs/specs/ and tests their acceptance criteria
  • Generates action/expected format steps executable through the UI
  • Covers edge cases, error scenarios, and regression checks

/open-pr

Create or update pull requests with rationale-focused descriptions.

  • Runs precondition checks (tests, lint, branch state)
  • Chains to /commit if uncommitted changes exist
  • Auto-detects ticket IDs from branch names (e.g., feature/ENG-123-description)
  • Generates QA steps via /generate-qa-steps
  • Creates PR with rationale-focused description using built-in template
  • Confirms before creating

/reflect

Run a structured retrospective on the current chat session.

  • Reviews every exchange for mistakes, unclear outputs, and wasted steps
  • Summarises what worked well and what caused friction
  • Proposes concrete improvements categorised as behaviour change, skill improvement, or new skill
  • Audits any skills used during the session: checks for success criteria, conciseness, and edge case handling
  • Detects repeated task patterns and suggests turning them into reusable skills
  • Proactively suggests a retro when it notices corrections, repeated clarifications, or frustration

/skill-creator

Create new skills, run evaluations, and iteratively improve skill quality.

  • Guides you from intent capture through draft, eval, and iteration
  • Runs test prompts in background and collects quantitative benchmarks
  • Generates eval review reports with eval-viewer/generate_review.py
  • Optimizes skill description for better triggering accuracy
  • Supports variance analysis across benchmark runs
  • Works for both new skills from scratch and improving existing ones

/test-in-browser

Test UI in the browser using the Playwright CLI.

  • Testing-focused workflow guide (open, snapshot, interact, verify)
  • Viewport testing across desktop, tablet, and mobile sizes
  • Visual verification via screenshots
  • Console and network inspection for debugging
  • Test code generation from CLI sessions
  • References /browse for full command documentation

/ticket

Create, read, update, and comment on Linear tickets from natural language.

  • Creates tickets with structured format (Context, Expected Behavior, Acceptance Criteria)
  • Reads ticket details by ID or lists assigned tickets
  • Adds comments to existing tickets
  • Updates status, assignee, priority, and labels
  • Maps to Linear fields via the official Linear MCP server
  • Requires one-time MCP setup: claude mcp add --transport http linear-server https://mcp.linear.app/mcp

/timesheet

Reconstruct and record weekly timesheets from calendar, Slack, Granola, and email data.

  • Pulls data from Google Calendar, Slack, Granola meeting notes, and Gmail in parallel
  • Categorises time into client projects and internal allocations based on attendees, email domains, and meeting titles
  • Learns and persists domain-to-client mappings between runs
  • Rounds durations to the nearest half hour, targets 40hrs/week by default
  • Handles overlapping meetings, stat holidays, PTO, and overtime
  • Writes per-day hour breakdowns to a markdown timesheet file with notes

Customization

Add sections to your project's CLAUDE.md to customize skill behavior. All overrides are optional -- skills fall back to sensible defaults.

Commit conventions

Override the default conventional commit format:

## Commit conventions

Use the format: [PROJ-XXX] type: description
Types: feature, bugfix, hotfix, chore
Always reference the ticket number in the footer.

PR template

Override the default PR description structure:

## PR template

Use this structure for PR descriptions:
- Title: [PROJ-XXX] Short description
- Include a "Testing" section with manual test steps
- Always link the Linear ticket

PR tools

Override the default GitHub CLI tool:

## PR tools

Use GitLab MR instead of GitHub PR.
Command: glab mr create

Design docs

Tell /address-pr-feedback where to find design documentation:

## Design docs

Design documents are stored in docs/decisions/.
Use ADR format (Architecture Decision Records).

Browser testing

Override default browser automation settings:

## Browser testing

- Base URL: http://localhost:3000/
- Browser: chrome
- Before browsing, start the dev server: `npm run dev`

Ticket conventions

Configure default Linear settings for /ticket:

## Ticket conventions

- Team: Engineering
- Default labels: bug, feature, improvement
- Project: Q1 2026 Roadmap
- Cycle: current
- Priority: 3 (Normal) unless specified

Override reference

CLAUDE.md Section Affects Default
## Commit conventions /commit format Conventional Commits
## PR template /open-pr description structure Built-in template
## PR tools /open-pr CLI tool gh (GitHub CLI)
## Design docs /address-pr-feedback doc location docs/plans/, docs/design/, docs/adr/
## Browser testing /browse, /test-in-browser Headed + persistent, Chromium
## Ticket conventions /ticket Ask user for team; priority 3 (Normal)

License

MIT

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors