The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation provides instructions to install the Bun runtime from bun.sh and the @vercel/agent-browser utility from the NPM registry.
[REMOTE_CODE_EXECUTION]: The skill recommends installing dependencies using a piped shell script (curl to bash) from the bun.sh domain.
[COMMAND_EXECUTION]: The skill utilizes shell execution via Node.js subprocesses to control browser interactions and manage automation workflows. It also provides configuration for an automated mode using the 'SKIP_PERMISSIONS=1' environment variable, which enables the agent to execute actions without explicit user intervention.
[PROMPT_INJECTION]: The agent is designed to navigate and capture snapshots of live social media sites like X.com and LinkedIn, processing this untrusted data through an LLM. This creates a surface for indirect prompt injection attacks where malicious content on a webpage could influence the agent's behavior.
Ingestion points: The agent-browser snapshot command reads content from live websites directly into the agent context.
Boundary markers: The provided examples do not include explicit delimiters or safety instructions to distinguish site content from system commands in the agent's prompt.
Capability inventory: The skill has broad capabilities including shell command execution, authenticated browser actions such as clicking and posting, and persistent file-based logging.
Sanitization: No explicit sanitization or filtering of extracted DOM content is demonstrated in the provided implementation snippets.

locoagent-social-media-automation