The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill establishes a connection to a remote Model Context Protocol (MCP) server at "https://unbrowse.ai/mcp/detik.com". This connection is used to provide the agent with the specific search and retrieval capabilities defined in the skill. The endpoint is managed by the skill's vendor.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from the external news site detik.com. This creates a potential surface for indirect prompt injection where malicious instructions embedded in news content could attempt to influence the agent.
Ingestion points: Data is pulled from detik.com through tools like detik_com__read_page and detik_com__get_search.
Boundary markers: There are no explicit delimiters or specific "ignore" instructions defined to wrap the ingested content.
Capability inventory: The skill possesses the ability to search, fetch web pages, and perform tasks through the Unbrowse MCP toolset.
Sanitization: No explicit content sanitization or validation logic is specified in the instruction set.
[COMMAND_EXECUTION]: The documentation includes shell commands, such as "claude mcp add", which are provided to the user for manual configuration of the MCP environment. These are standard installation and setup instructions and are not executed automatically by the skill.

detik