perplexity-agent-skill

Perplexity Web Research Skill

An Agent Skills package that leverages the Perplexity CLI for web research, deep reasoning, and independent validation. Compatible with any AI coding assistant that supports the Agent Skills standard (Claude Code, GitHub Copilot, Cursor, Windsurf, and 30+ other agents).

⚠️ Requirements

You need a Perplexity API key to use this skill.

• Get your API key at: https://www.perplexity.ai/account/api/keys
• Perplexity Pro subscribers get $5 in free API credits each month 💰
• Pay-as-you-go pricing available for non-subscribers

What This Skill Does

This skill enables AI assistants to use Perplexity's specialized models for three primary purposes:

1. Current or External Knowledge

• Documentation lookups and API references
• Library versions and recent technology changes
• Error message research
• Current events and fact checking

2. Independent Second Opinions

• Architecture and design decision validation
• Technology choice comparisons
• Alternative perspectives on problem-solving approaches
• Pre-implementation validation

3. Deep Reasoning Delegation

• Complex multi-step problem solving
• Algorithmic optimization and system design
• Trade-off analysis with many variables
• Mathematical and quantitative reasoning

Why CLI over MCP?

This skill uses the Perplexity CLI (llm with llm-perplexity plugin) instead of MCP servers because:

• Lower token usage: CLI results are more concise than MCP server overhead, preserving valuable context window space
• Faster execution: No server startup/connection delays - just direct bash command execution
• Simpler integration: Straightforward CLI tool that works anywhere bash is available
• Complementary approach: Works alongside standard web tools (WebFetch, WebSearch) rather than replacing them

Installation

Quick Install (Recommended)

Use the Vercel Skills CLI to install for any compatible agent:

# For GitHub Copilot
npx skills add xpepper/perplexity-agent-skill -a github-copilot

# For Claude Code
npx skills add xpepper/perplexity-agent-skill -a claude-code

# For Cursor
npx skills add xpepper/perplexity-agent-skill -a cursor

# Or install for all agents
npx skills add xpepper/perplexity-agent-skill

The skill will be automatically installed to the appropriate location for your agent and ready to use immediately.

Manual Installation

Alternatively, install manually by copying the skill directory:

For VS Code / GitHub Copilot:

git clone https://github.com/xpepper/perplexity-agent-skill.git
cp -r perplexity-agent-skill/perplexity-web-research ~/.copilot/skills/

For Claude Code:

git clone https://github.com/xpepper/perplexity-agent-skill.git
cp -r perplexity-agent-skill/perplexity-web-research ~/.claude/skills/

For other agents: Check your agent's documentation for the skills directory location.

Prerequisites

1. Perplexity API Key (Required)

You must have a Perplexity API key. Get yours at:

• API Settings: https://www.perplexity.ai/settings/api
• Perplexity Pro subscribers: Get $5 in free API credits each month
• Pay-as-you-go: Available for all users

2. Install llm CLI with Perplexity Plugin

# Install llm CLI
pip install llm
# Or using Homebrew
brew install llm

# Install the Perplexity plugin
llm install llm-perplexity

# Configure your Perplexity API key
llm keys set perplexity
# When prompted, paste your API key from https://www.perplexity.ai/settings/api

3. Verify Setup

llm -m sonar 'hello'

If you see a response from Perplexity, you're all set! ✓

For detailed setup instructions and troubleshooting, see perplexity-web-research/references/setup.md.

Usage

The skill automatically triggers when the AI assistant needs:

• Current documentation or up-to-date information
• Independent validation of decisions or approaches
• Complex reasoning or analytical thinking delegation

It can be invoked directly (inline Perplexity queries) or spawned as a subagent for focused research.

Model Selection

The skill uses different Perplexity models based on the task:

Documentation

• Full skill documentation: perplexity-web-research/SKILL.md
• Implementation plan: specs/perplexity-skill-plan.md
• Setup guide: perplexity-web-research/references/setup.md
• llm CLI: https://llm.datasette.io/
• llm-perplexity plugin: https://github.com/hex/llm-perplexity

Development

Packaging the Skill

After making changes to the skill, regenerate the .skill package:

./package-skill.sh

This script:

• Validates the skill directory structure
• Removes the old package
• Creates a new zip archive with .skill extension
• Excludes system files (.DS_Store, pycache, etc.)

License

MIT License - see LICENSE file for details.

Copyright (c) 2026 Pietro Di Bello

Contributing

Contributions are welcome! Please feel free to submit issues or pull requests.