olo-dot-io/uni-cli/unicli-browser
Claude Code and agent skill by olo-dot-io/uni-cli
- community
Checking that the SKILL.md file is still in the repository.
Control broker-owned hidden, existing-Chrome, or remote browser targets with explicit Agent identity, visibility, profile partitions, and lifecycle.
Install
npx vskill@latest install olo-dot-io/uni-cli/unicli-browserSKILL.md
When to Use
Use unicli browser for browser lifecycle, page interaction, session sharing, target handoff, and diagnostics. unicli operate is a compatibility alias for the page-action subset.
Provider Decision
| Requirement | Provider and visibility | |
|---|---|---|
| No window; default automation | managed + hidden (default) | |
| Existing signed-in Chrome tab, no activation | chrome + background | |
| Explicit user-visible Chrome interaction | chrome + foreground | |
| Explicit cloud/browser service | remote + hidden |
No provider falls back to another. Managed and remote providers cannot become foreground. Existing Chrome cannot promise hidden operation.
Quick Start
Agent Sessions and Reuse
A stable Agent session owns targets across turns. Different Agent sessions use different targets, while a shared profile partition can reuse login/storage.
Use --profile-partition team-login to share storage deliberately. Use --isolated for a disposable browser context inside that partition. --workspace is a compatibility spelling for the same partition concept.
Existing Chrome Without Foreground Activation
Install the native host and extension once, keep an existing normal Chrome window open, list tabs, then claim a specific tab:
Background allocation must preserve focused window and active-tab state. If Chrome cannot satisfy that postcondition, the command fails instead of opening or focusing a window. Use --focus only for an explicit foreground request.
Broker Lifecycle
The broker is one authenticated, owner-only local control plane. Starting it does not start Chrome. Browser processes are provider-owned and lazy. Session TTL and target leases are visible through browser status and browser sessions.
Authentication and Profiles
Managed runtimes use Uni-CLI-owned profiles under ~/.unicli/; they never run CDP against Chrome's default user-data-dir. A selected local profile seeds the automation profile. --ephemeral explicitly selects an empty profile.
Chrome 136+ ignores remote-debugging switches on its default data directory. RemoteDebuggingAllowed=false blocks even Uni-CLI automation profiles until an administrator removes the policy or sets it true. The policy does not make the default-profile CDP path supported.
Remote Browser
Status redacts credentials, paths, query tokens, and headers. Malformed remote configuration is a hard error and never falls back locally.
Page Workflow
openthe URL.stateto obtain current refs.- Use
click,type,select, orkeyswith those refs. - Run
stateagain after navigation or DOM replacement. - Prefer captured JSON APIs when
networkreveals a stable endpoint.
Useful commands:
Diagnostics
unicli browser doctor --json is the routing truth:
default_path: managed/hidden availability and profile source.broker: stopped, running, or exact endpoint error.providers: managed runtimes, Chrome native-host state, redacted remote state.sessions: live Agent turns, target leases, and tombstones.checks[*].next_step: exact repairs.
unicli browser doctor --repair starts only the broker. It never starts a browser provider and never creates an about:blank placeholder target.
| Problem | Exact next action | |
|---|---|---|
| Broker unavailable | unicli browser broker restart | |
| Managed browser unavailable | Install Chrome or use explicit --ephemeral | |
| Chrome provider unavailable | Install native host/extension and keep a normal Chrome window open | |
| Remote config invalid | Correct or unset UNICLI_CDP_ENDPOINT / UNICLI_CDP_HEADERS, restart broker | |
| Exit 77 | Restore auth with explicit profile/cookie flow |
Security
- Security scanPassed, 93 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/unicli-browser/SKILL.md
Popularity
- Repository stars
- 269
- Repository forks
- 3
- 7-day trend score
- 28
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/olo-dot-io/uni-cli/unicli-browser)