The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches a setup script for the Bun runtime from bun.sh, a well-known service. The script's integrity is verified using a hardcoded SHA256 checksum before execution.
[COMMAND_EXECUTION]: Executes the downloaded Bun installer script via the shell to set up the execution environment.
[DATA_EXFILTRATION]: Facilitates remote access to the browser session through an ngrok tunnel. While this represents a significant network exposure surface, the skill includes explicit security warnings and requires user consent before enabling the tunnel.
[INDIRECT_PROMPT_INJECTION]: As the skill allows a remote agent to read and interact with arbitrary web content, it creates an attack surface for indirect prompt injection from malicious websites.
Ingestion points: Paired browser tabs reading HTML, text, and snapshots from external web pages.
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for the content read from the web.
Capability inventory: Navigating URLs, clicking elements, filling forms, taking screenshots, and executing JavaScript via eval.
Sanitization: No specific sanitization or validation of the external web content is described.

pair-agent