Use Parallel Skills for Web Search and Content Extraction

Who It's For

If you want an agent to search the web, extract page content, conduct deep research, or enrich data, rather than relying only on model memory, these Parallel skills are for you. The official description presents them as Agent Skills for Parallel, covering web search, content extraction, deep research, and data enrichment.

Most skills that actually initiate calls require parallel-cli to be installed and logged in, and the account to be funded. If any one of these is missing, execution-oriented skills cannot proceed. For a first practice, start with search or extraction: the public pages provide copyable invocation syntax, and it is easier to compare the returned result to determine whether the run worked. Deep research has an example, but it is a heavier task and is better saved until you have confirmed that the account has a balance. Data enrichment is named as a capability, but this page has no corresponding command, so it cannot be treated as a demonstrated operation.

Do not infer language coverage from examples such as React 19, and do not interpret any one agent as being exclusively supported. This guide uses only the official parallel-cli skill installation command.

Complete Your First Attempt in 3 Minutes

Use only the following official installation command to install all Parallel skills:

parallel-cli skills install

Prerequisites: parallel-cli must already run on the machine, and the command line must be logged in with a funded account. If installation, login, balance checking, and skill installation are not complete, the official instructions tell you to run /parallel:parallel-cli-setup in the agent, which guides you through the process. This page does not say which directory the skill files should be copied to, so do not guess the path yourself.

After installation succeeds, first perform a check that does not access the external network: confirm that parallel-cli still runs and that the agent can recognize /parallel:parallel-cli-setup. Once both are confirmed, you can treat the installation side as complete.

Actual searches access external webpages and consume funded credits. First ask the agent to repeat the search terms it is about to run; after you confirm them, run the official example:

/parallel:parallel-web-search latest React 19 features

Success signal: the agent returns search results related to the new features of React 19, rather than generic advice. The public description does not specify billing details or a free quota; it only requires a funded account. This practice verifies only that search results can be returned.

Use Cases

Case one: perform a web search using the official example. With parallel-cli logged in and the account funded, use /parallel:parallel-web-search and replace the search terms with the topic you actually need to investigate. The official example is latest React 19 features. Confirm the search terms first, then execute. This page provides no filtering, result-count, or language parameters, so the instructions cover only starting the search and viewing the response.

Case two: extract the content of a public page. The official example is:

/parallel:parallel-web-extract https://docs.parallel.ai

This visits the URL and extracts its content. Before running it, confirm that this is the page you intend to extract. The public materials do not specify extraction length, format, or failure retries, and you should not fold deep research or data-enrichment steps into it. Deep research has a separate example, /parallel:parallel-deep-research competitive landscape of AI code assistants, but that is a more intensive research task and is not part of this case.

Boundaries of Use

Most execution-oriented skills depend on parallel-cli being installed, logged in, and funded. Do not interpret a failure while you are not logged in or funded as evidence that the skill itself is unavailable. Data enrichment is named as a capability, but this page has no corresponding command, so it cannot be treated as a demonstrated operation.

The only installation command is parallel-cli skills install. The public materials provide no uninstall steps, and they do not generalize language coverage from the React 19 example.

Search, extraction, and deep research all access the external network and use account credits, so confirm before executing. Repository URL:

https://github.com/parallel-web/parallel-agent-skills