Lightcone Browse

Computer-use automation via Lightcone. Activate when user asks to browse websites, scrape web pages, automate web tasks, check prices, monitor sites, fill fo...

Security Scan

The skill's requirements and runtime instructions are coherent with a cloud-based browser automation tool, but the API key grants broad access to remote browsing/screenshot data and you should verify the vendor and scope of the key before use.

Purpose & Capability

The skill describes cloud browser automation and only requests a single API key (TZAFON_API_KEY), which is consistent with calling a third‑party browsing service. Note: the env var name (TZAFON_API_KEY) is somewhat unusual compared to the skill name (Lightcone); confirm this is the documented key name for the Lightcone service or the intended provider.

Instruction Scope

SKILL.md instructs the agent to create cloud browser sessions, take screenshots, and return page content — all expected for this capability. This means page HTML and screenshots will be sent to the remote service; that is expected behavior but is privacy-sensitive and could expose credentials or personal data if used on authenticated pages.

Install Mechanism

Instruction-only skill with no install spec and no code files, so nothing is written to disk by an installer. Low technical installation risk.

Credentials

Only one environment variable (an API key) is required, which is proportionate to a cloud browsing service. However, that key likely allows creating remote machines and viewing all pages/screenshots — a high-privilege capability. Ensure the key is scoped appropriately and dedicated to non-sensitive tasks.

Persistence & Privilege

The skill is not marked always:true and has no special persistence or system-wide config changes. Be aware the platform default allows autonomous invocation; combined with the API key this permits the agent to invoke remote browsing without further user prompts.

Assessment

This skill appears to do what it says (cloud browser automation), but it will send screenshots and page content to the remote service using the provided API key. Before installing: verify the vendor (docs.lightcone.ai) and that TZAFON_API_KEY is the correct/documented credential name; use a dedicated, least-privilege API key; avoid running this skill on pages with sensitive credentials or personal data; review the service's privacy/security policy; monitor API key usage and rotate/revoke the key if you see unexpected activity. If you need higher assurance, ask the publisher for details about what the API key can do and whether it can be scoped or rate-limited.

Delegate browser tasks to a cloud computer powered by Lightcone. Northstar sees the screen, decides what to click, type, or scroll — and acts. You describe what you want done, Northstar does it.

When to use Lightcone

• No local browser available (headless, Docker, server deployments)
• Need parallel sessions (multiple sites at once)
• Vision-based automation (Northstar reads the screen — no selectors, no scripts to maintain)
• Desktop or browser environments

Tool: lightcone_browse

Parameters

Parameter	Required	Description
instruction	Yes	Plain language task for Northstar
url	No	Starting URL to navigate to first
maxSteps	No	Max steps before stopping (default: 50)

Example instructions

Price checking:

instruction: "Find the current price of a MacBook Air M4 on Amazon"
url: "https://www.amazon.com"

Data extraction:

instruction: "Go to the pricing page and extract all plan names, prices, and feature lists"
url: "https://example.com/pricing"

Form filling:

instruction: "Fill out the contact form with name 'Jane Doe', email '[e-mail]', and message 'Hello'"
url: "https://example.com/contact"

Tips for good instructions

• Be specific about what "done" looks like: "find the price" not "look around"
• Name the exact elements: "click the Sign In button" not "log in somehow"
• For multi-step tasks, describe the sequence: "First search for X, then sort by price, then get the top 3 results"
• Include what to extract: "tell me the title, price, and rating"

How it works

1. Lightcone creates a cloud computer with a browser
2. Northstar sees the screen via screenshots
3. It decides what to click, type, or scroll — step by step
4. When done, it returns the page content and a screenshot
5. The computer is automatically shut down

For step-by-step manual control, see the lightcone-session skill.