The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill connects to a remote Model Context Protocol (MCP) server hosted at unbrowse.ai. This endpoint serves as a proxy to interact with hex.pm and is a vendor-managed resource provided by the skill author.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from the external hex.pm registry. This creates an attack surface where maliciously formatted package metadata or descriptions could attempt to influence the agent's behavior.
Ingestion points: Data is retrieved via the hex_pm__get_packages and unbrowse.scrape tools which return text and links from the target site.
Boundary markers: The instructions do not define specific delimiters to wrap external content, though they advise the agent to only report successful results.
Capability inventory: The skill facilitates remote web interactions through the vendor's infrastructure but does not provide tools for local file system modification or direct command execution on the host machine.
Sanitization: The skill relies on the remote MCP server to return verified results from the site's API, minimizing raw scraping risks.

hex-pm