The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to run the Midscene CLI and perform connectivity prechecks on local ports (9222 and 3766) to detect available browser instances.
[EXTERNAL_DOWNLOADS]: Fetches and executes the @midscene/web package directly from the npm registry using npx -y to perform browser automation tasks.
[DATA_EXFILTRATION]: Manages sensitive model configuration and API keys through environment variables. It correctly advises users to store these secrets in .env files to prevent exposure.
[PROMPT_INJECTION]: As a browser automation tool, the skill ingests data from external, untrusted web pages via screenshots and element analysis.
Ingestion points: External websites accessed via connect --url or within the browser session.
Boundary markers: None explicitly defined in the prompt instructions for external data.
Capability inventory: Ability to execute bash commands, control browser sessions (click, type, navigate), and write report files to the local filesystem.
Sanitization: Relies on the underlying Midscene.js processing and the agent's internal logic to interpret visual data.

browser-automation