Skip to content
Get packs

Guide · Agent skill

How to use the OpenCLI Usage Guide agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The opencli-usage skill is a set of instructions that teaches an AI agent how to work with the OpenCLI command-line tool. OpenCLI packages website integrations (called adapters) and direct Chrome browser controls into commands the agent can search, inspect and run. This guide is for people who already have OpenCLI and want the agent to use it correctly; after reading it you will know how to install the skill, how to ask for things in a way that triggers it, and what to watch for.

OpenCLI Usage Guidejackwener/OpenCLI · Apache-2.0

Quick start

  1. 1Install OpenCLI. Make sure the OpenCLI command-line tool (opencli) is installed, since the skill only documents an existing tool.
  2. 2Download the skill ZIP. Download the opencli-usage ZIP from this page.
  3. 3Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder.
  4. 4Or upload in the Claude apps. Alternatively, upload the ZIP under Settings → Capabilities → Skills.
  5. 5Connect Chrome for browser tasks. For browser commands, connect Chrome to OpenCLI and check it with opencli browser chrome describe or doctor.
  6. 6Ask in plain language. Describe a task involving OpenCLI, websites or Chrome, and the agent will use the skill.
  7. 7Review before adapter tests. Check adapters with adapter check before running adapter test, since tests can perform real writes.

What the opencli-usage skill makes the agent do

An agent skill is a folder with a SKILL.md file that an agent reads when a task matches the skill's description. This skill covers how to discover and invoke OpenCLI capabilities for websites and Chrome.

Once installed, the agent searches for commands instead of loading the whole catalog, reads JSON results and error codes, opens and closes browser sessions, and checks adapter and login status. For browser interaction and for writing adapters, the skill points to two other OpenCLI skills, opencli-browser and opencli-adapter-author.

  • Search commands by name or description with opencli list [query] instead of dumping every command into context.
  • Read results as JSON envelopes ({ok:true,data:...}) and recover from error codes.
  • Open, inspect and close browser sessions tied to a --session name.
  • Check adapter status, adapter tests and website login status.

Install the skill on Claude

You need the OpenCLI command-line tool (opencli) installed first; the skill only documents an existing tool. Then download the opencli-usage ZIP from this page and install it in one of two ways.

For a personal install, unzip the file into ~/.claude/skills/. To limit it to one project, unzip it into that project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

  • Personal: unzip into ~/.claude/skills/ so it is available in all your projects.
  • Project only: unzip into the project's .claude/skills/ folder.
  • Claude apps: upload the ZIP under Settings → Capabilities → Skills.

Trigger the skill with natural requests

You do not need a special command to use the skill. Describe the task in plain language that involves OpenCLI, websites or Chrome, and the agent will match it to the skill's description: find commands, inspect schemas, interpret results, or choose between browser and adapter tools.

If you want to check what the skill exposes, you can run the bundled listing yourself. The source shows skills list and skills read opencli-browser -f plain for reading bundled skill content.

  • "Find an OpenCLI command that searches GitHub repositories."
  • "Open example.com in a research browser session, then show me the page state."
  • "Check whether my Chrome profile is connected to OpenCLI and diagnose it if not."
  • "Read the last JSON error from that opencli run and tell me what to try next."
opencli list search opencli list --site github --limit 20 opencli github search --help -f json

Example: browser sessions and adapter checks

Page commands use --session to keep a browser session separate. The agent can open a page, inspect its state and close the session when done. Native Chrome methods need no session. --profile selects a connected Chrome profile, and profile use <name> sets the default. Use doctor for connection problems.

For adapters, the agent can initialise one, check it, test it with arguments, and view status. Be aware that adapter test runs the real command and can perform writes.

  • Use --profile to choose a connected Chrome profile; profile list shows what is available.
  • Use adapter check before adapter test to validate the adapter first.
  • adapter eject makes an official site adapter editable locally; adapter reset removes that local override.
opencli browser open https://example.com --session research opencli browser state --session research opencli browser close --session research opencli browser chrome describe opencli adapter init example/search opencli adapter check example/search opencli adapter test example/search --args '{"query":"hello"}' opencli adapter status

Safety notes: scripts, network and logged-in sessions

This skill runs scripts and uses the network. It allows the Bash(opencli:*) tool and Read. The agent can act inside logged-in sessions, so it may see or change data in accounts that your Chrome profile is signed into.

Adapter tests execute real commands and can perform writes. The source also says that when a write outcome is unknown, the agent should inspect state before repeating the action. Avoid using this skill with sensitive accounts or shared browser profiles, and review what an adapter test will do before running it.

  • Do not use it with sensitive accounts or shared browser profiles.
  • Adapter tests can perform real writes; check the adapter first.
  • Unknown write outcomes: inspect state before repeating an action.
  • Website logins are needed for adapters that use them; auth status reports login facts and auth refresh touches sessions.

Tips for getting reliable results

Read command help before running an unfamiliar command. The skill says to inspect exact inputs using help, and to use -f json on help when you want structured output. Paginated results include nextOffset when there is more; pass it with --offset to continue.

An empty successful result is not the same as a failure, so do not treat it as an error. Explicit formats such as -f table, yaml, csv, md or plain are available only where a command's help lists them.

  • Use opencli github search --help -f json to see a command's inputs in structured form.
  • Continue paginated results with --offset <nextOffset>.
  • Logs go to stderr, so separate them from JSON on stdout when parsing.
  • Use daemon status to check the bridge process if browser commands fail.

What people use it for

  • Finding the right OpenCLI command for a website, such as searching GitHub with opencli list --site github.
  • Reading data from a site through a connected Chrome profile using browser sessions.
  • Automating a repeated website task by initialising, checking and testing a reusable adapter.
  • Diagnosing a failed run by reading the JSON error code and hint, then choosing the next step.

Frequently asked questions

Is the OpenCLI usage skill free?

The skill is licensed under Apache-2.0, as stated on this page. The source does not cover pricing for OpenCLI or the skill beyond that, so check the upstream repository for current terms.

Do I need OpenCLI installed before using this skill?

Yes. The skill only documents an existing command-line tool, so you need opencli installed first. Without it, the skill has nothing to run.

Where do I put the skill folder for Claude?

Unzip the downloaded ZIP into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder for one project. You can also upload it in the Claude apps under Settings → Capabilities → Skills.

Does the skill need Chrome?

Only for browser commands. Those need Chrome connected to OpenCLI, and adapters that use websites need you to be logged in to those sites.

Is it safe to let an agent use OpenCLI on my accounts?

Be careful. The skill runs scripts, uses the network and can act inside logged-in sessions. Adapter tests can perform real writes, so avoid sensitive accounts and shared browser profiles.

How is this different from opencli-browser or opencli-adapter-author?

This skill covers finding commands, reading results and choosing tools. The opencli-browser skill covers browser interaction and tab management, and opencli-adapter-author covers writing adapters.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits developers who already use OpenCLI and want an agent to drive it correctly, and people who automate repeated website tasks through reusable adapters. It is not useful without OpenCLI installed, and it is a poor fit for sensitive accounts or shared browser profiles, where an agent acting in logged-in sessions carries real risk.

Download the skill