The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch and process arbitrary text from external web pages on aramco.com, which could contain instructions meant to influence the agent's behavior.
Ingestion points: Data enters the agent's context through tools like aramco_com__read_page and aramco_com__get_article as described in SKILL.md.
Boundary markers: There are no explicit instructions or delimiters provided to help the agent distinguish between the site's data and potential malicious instructions embedded in that data.
Capability inventory: The skill includes powerful tools such as unbrowse.run, unbrowse.browse.act, and unbrowse.browse.open which allow the agent to perform actions and navigate the site based on the information it reads.
Sanitization: The skill instructions do not mention any sanitization, filtering, or validation of the content retrieved from the external site before processing.
[EXTERNAL_DOWNLOADS]: The skill connects to a remote MCP server hosted at https://unbrowse.ai/mcp/aramco.com as defined in server.json and SKILL.md. This is a vendor-owned resource used for the skill's core functionality.
[REMOTE_CODE_EXECUTION]: The skill relies on a remote HTTP-based Model Context Protocol (MCP) server to execute its primary functions, meaning the logic for the tools is executed on the vendor's infrastructure.

aramco