- Home
- Skills
- Simota
- Agent Skills
- Navigator
navigator_skill
- Shell
8
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill simota/agent-skills --skill navigator- SKILL.md8.8 KB
Overview
This skill automates one-off browser tasks using Playwright and Chrome DevTools to reliably complete a specified objective. It performs navigation, form interactions, data extraction, network and console monitoring, and captures evidence (screenshots, HAR, video) to prove outcomes. The focus is task completion and accurate reporting rather than producing reusable test code. Safety, explicit waits, and artifact preservation are built into every run.
How this skill works
Navigator connects to a Playwright MCP server and/or CDP to control a browser context, evaluate page state, and execute deterministic interactions. It runs through five phases—Recon, Plan, Execute, Collect, Report—identifying selectors, handling auth, performing actions, capturing logs/HAR/screenshots, and packaging results. It watches console and network activity, saves storage state when needed, and emits a concise execution summary and artifacts for downstream agents.
When to use it
- Collect structured data from a site or paginated UI when accuracy matters
- Reproduce and capture evidence for a bug or visual issue (screenshots/video/HAR)
- Automate form completion or multi-step flows once (non-test one-off tasks)
- Verify a deployment or quick production check that needs browser context
- Perform security or performance validation with network and console traces
Best practices
- Always confirm target URL, scope, and authentication method before running
- Use explicit element waits and avoid fixed timeouts; retry transient failures
- Ask for permission before destructive actions, credential entry, or PII collection
- Capture screenshots after key steps and include HAR/console logs for failures
- Persist session storage state via context.storageState() instead of hardcoding secrets
- Limit scraping volume (ask before >100 pages) and respect rate limits and ToS
Example use cases
- Log into a dashboard, export a report table to JSON/CSV, and save screenshots for audit
- Reproduce a checkout bug, record a short video, collect HAR and console errors, and hand off artifacts to triage
- Fill and submit a multi-step form, verify the confirmation page, and capture before/after evidence
- Crawl paginated product listings, normalize fields, and deliver structured data to a builder agent
- Run a quick performance capture for a slow page and forward HAR and metrics to a tuning agent
FAQ
No. Navigator focuses on executing and completing tasks directly; it does not produce reusable test suites.
How are credentials handled?
Credentials must be provided via environment variables or saved session state. Never hardcode secrets or store them in plaintext.