A collection of skills for AI coding agents. Skills are packaged instructions and references that extend agent capabilities.
Skills follow the Agent Skills specification.
Find similar products using dupe.com's visual search API.
Use when:
- User provides a product URL and asks for similar/alternative products
- User shares an image URL and wants to find matching items
- Searching for dupes, alternatives, or look-alikes
Features:
- Product URL search via dupe.com API
- Direct image URL search capability
- Returns matches with prices, store info, and direct links
- Provides similarity scores and product rankings
- Generates dupe.com result page links
Launch AI-powered product research tasks from your agent — the same pipeline that powers dupe.com.
Use when:
- Researching the best products in a category ("best wireless earbuds under $100")
- Comparing products with specific priorities (value, quality, reviews, etc.)
- Getting AI-powered product recommendations with detailed reasoning
- Generating a comprehensive research report for a product category
Features:
- 5 research archetypes (deal hunter, quality-first, quick pick, social proof, deep research)
- Real-time progress tracking across 9 research phases
- Returns product candidates with pricing, images, and brand info
- Generates full editorial research articles
- Links to published report pages on dupe.com
- Protected by BOTCHA — agent solves a speed challenge to authenticate (no API keys or SDKs needed)
Error prevention and best practices enforcement for AI agent-assisted coding.
Use when:
- Writing new code or functions
- Refactoring existing code
- Implementing API integrations
- Handling user input or external data
- Working with async operations
- Reviewing AI-generated code
Features:
- 10 core error prevention principles
- Input validation, null safety, async handling
- TypeScript strict mode patterns
- Biome migration guide (ESLint → Biome)
- AI code review checklist
- Model configuration recommendations (Claude 3.5+, GPT-4+, extended thinking)
- 60KB+ of reference documentation
Address GitHub PR review feedback end-to-end from your agent.
Use when:
- User types
/reviewsor asks to "address review comments" - Going through inline PR feedback after a code review
- Responding to a specific PR by number, URL, or branch name
Features:
- Fetches all reviews, inline comments, and conversation threads via REST + GraphQL
- Triages actionable vs ambiguous items; skips already-resolved threads
- Applies code changes, posts replies, and resolves threads incrementally
- Surfaces ambiguous items and disagreements for human review before proceeding
- Posts a structured summary comment to the PR
- Never pushes — the human always reviews the commit and pushes
Rebase the current branch onto origin/main and resolve conflicts intelligently.
Use when:
- User types
/rebasedor asks to "rebase and fix conflicts" - Branch has fallen behind main and needs to be brought up to date
- A rebase is already in progress and needs help resolving remaining conflicts
Features:
- Pre-flight checks for dirty working tree, in-progress rebases, and edge cases
- Merges both sides of each conflict by understanding intent, not blindly picking one
- Regenerates lock files (
bun.lockb,package-lock.json, etc.) rather than hand-merging - Works commit-by-commit through multi-step rebases
- Never force-pushes without explicit user confirmation
Generate ASCII art from images or text using shape vector rendering.
Use when:
- Creating ASCII text banners for terminals, READMEs, or CLI tools
- Converting images to ASCII art with high-quality edge detection
- Building terminal splash screens or headers
- Generating retro/terminal aesthetic text
Features:
- Text-to-ASCII with customizable fonts (Arial Black, Impact, etc.)
- Image-to-ASCII conversion with sharp edge detection
- Shape vector algorithm for superior quality vs traditional brightness mapping
- Adjustable contrast, size, and invert options
- Built-in demo mode
Pin a short personal reminder to the Claude Code status line.
Use when:
- You type
/note <reminder>(e.g./note revert the test-only flag before merging) - You say "remind me to…", "note to self", "pin a note", "set a reminder"
- You want to clear it:
/note clear
Features:
- Synthesizes your input into a terse one-line reminder (strips filler)
- Renders on its own status-line row as a bold black-on-yellow sticky note
- Scoped to the current session — notes don't leak across Claude sessions
- Persists across compactions/restarts until cleared
- Ships an idempotent installer that wires the status line for you
Install any skill with a single command via skills.sh:
npx skills add dupe-com/skills/dupe-research
npx skills add dupe-com/skills/dupe
npx skills add dupe-com/skills/nomistakes
npx skills add dupe-com/skills/ascii-renderer
npx skills add dupe-com/skills/reviews
npx skills add dupe-com/skills/rebased
npx skills add dupe-com/skills/noteThis works with Claude Code, OpenCode, Cursor, Windsurf, Cline, Codex, AMP, Copilot, and any agent that supports the agentskills.io standard.
git clone https://github.com/dupe-com/skills.git
cp -r skills/skills/<skill-name> ~/.claude/skills/Each skill follows a consistent structure:
skills/skill-name/
├── SKILL.md # Main instructions (required)
├── README.md # Detailed documentation
├── metadata.json # Version, author, tags
├── LICENSE # Skill-specific license
├── references/ # Additional documentation
└── scripts/ # Executable utilities
Use the template in template/SKILL.md as a starting point.
---
name: my-skill-name # lowercase-hyphenated
description: What it does and when to use it
---- Keep SKILL.md under 500 lines - Move detailed docs to
references/ - Write for agents - Clear, actionable instructions
- Include examples - Show both ❌ BAD and ✅ GOOD patterns
- Test your skill - Run
node scripts/validate-skill.cjsif available
- Fork this repository
- Create a new skill in
skills/your-skill-name/ - Follow the structure and guidelines above
- Submit a pull request
Individual skills may have their own licenses. See each skill's LICENSE file.
Repository structure and shared tooling: MIT
Built by Dupe.com - AI-powered shopping research.