A security-hardened Perplexity API integration for real-time web-grounded search with citations.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install perplexity-safe
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install perplexity-safe using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Perplexity AI Search (Safe Edition)?
This skill provides robust access to the Perplexity API, allowing AI agents to perform web-grounded searches directly from their workspace. Unlike standard LLMs that rely on training data, this Openclaw Skills integration provides up-to-date information, news, and factual data with verifiable source citations. It bridges the gap between static knowledge and the live web, making it an essential tool for any research-heavy workflow.
This specific implementation is a security-hardened version of the original bash-based search tools. It addresses critical vulnerabilities such as command injection by passing user input via environment variables rather than code interpolation. By using this within the Openclaw Skills ecosystem, developers can ensure that their agents interact with external APIs safely while maintaining high performance and data accuracy.
Perplexity AI Search (Safe Edition) Use Cases
- Fetching current news, stock prices, or recent technical developments that occurred after an LLM's cutoff date.
- Generating research-style answers with explicit source citations and references for academic or professional reports.
- Performing deep reasoning analysis using advanced models like sonar-reasoning for complex, multi-step queries.
- Validating facts against real-time web data to reduce hallucinations in agentic workflows using Openclaw Skills.
How Perplexity AI Search (Safe Edition) Works
- The user or agent triggers the
perplexity_search.shscript with a specific query and optional model parameters. - The skill performs strict input validation, checking numeric ranges, model names, and query length to prevent denial-of-wallet attacks.
- Authentication is handled securely; the skill retrieves the API key from a local config or environment variable and passes it to curl via a temporary protected file.
- The query is dispatched to the Perplexity API using the selected model (e.g., sonar or sonar-pro) and search context depth.
- The skill processes the API response and returns the web-grounded answer in the requested format (text, markdown, or JSON) for the agent to process.
Perplexity AI Search (Safe Edition) Setup
To integrate this search capability into your Openclaw Skills setup, follow these steps:
- Obtain an API key from the Perplexity developer dashboard.
- Option 1 (Recommended): Create a
config.jsonfile in the skill directory:
{
"apiKey": "pplx-your-key-here"
}
- Option 2: Export the key as an environment variable:
export PERPLEXITY_API_KEY="your-key-here"
- Ensure
bash(4.0+),curl, andpython3are installed on your system. - Test the installation by running a basic search command:
scripts/perplexity_search.sh "What are the latest Openclaw Skills updates?"
Perplexity AI Search (Safe Edition) Data Schema & Taxonomy
The skill manages configuration and API communication using the following data structures:
| Component | Purpose | Format |
|---|---|---|
config.json |
Stores the Perplexity API key and default settings | JSON |
Request Body |
Contains the query, model selection, and max tokens | JSON |
Output |
The final search result with citations and model metadata | Markdown / JSON |
Search Context |
Level of research depth (low, medium, high) | String Enum |
Perplexity AI Search (Safe Edition) Advanced Features
- Security-hardened architecture that prevents command injection by avoiding shell interpolation.
- Multi-model support including sonar, sonar-pro, and sonar-reasoning-pro for varying levels of research complexity.
- Enhanced privacy features that use temporary config files for curl to hide sensitive API keys from process listings.
- Configurable output formats (text, markdown, or JSON) to fit seamlessly into automated Openclaw Skills pipelines.
- Query length capping and strict model allowlists to manage API costs and prevent misuse.
SKILL.md
Loading
Related Openclaw Skills
Miniflux News
A robust tool to fetch, summarize, and manage unread RSS entries from your Miniflux instance via its REST API.
Write My Blog
An autonomous AI skill for creating, managing, and deploying full-featured blogs across multiple database architectures and hosting providers.
Podcast Discovery & Generation
An automated end-to-end pipeline to discover trending topics, research them, and generate fact-checked podcast episodes with professional audio.
Cricket Live
A comprehensive cricket tracking suite for OpenClaw agents providing live scores, IPL updates, and automated match alerts.
Message Injector
A powerful OpenClaw extension that automatically prepends custom instructions or context to every user message before the agent processes it.
Workspace Indexer
An automated utility that maintains a structured and searchable index of your workspace directories and project metadata.
Featured*
