The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation suggests installation using npx via the vendor's website at https://unbrowse.ai. It also establishes a remote connection to an external Model Context Protocol (MCP) server hosted at https://unbrowse.ai/mcp/creativecommons.org to provide tool definitions and logic.
[INDIRECT_PROMPT_INJECTION]: The skill provides tools like creativecommons_org__read_page and creativecommons_org__get_root that ingest data from the creativecommons.org website. This data is passed to the agent, creating an attack surface where malicious instructions embedded in the website could potentially manipulate agent behavior.
Ingestion points: Tools described in SKILL.md fetch content from external URLs on creativecommons.org.
Boundary markers: The instructions do not specify the use of delimiters or specific prompts to ignore instructions within the fetched content.
Capability inventory: The skill includes tools for web automation and task execution (unbrowse.run, unbrowse.browse.open, unbrowse.scrape).
Sanitization: No content sanitization or validation methods are described for the data retrieved from the target website.
[COMMAND_EXECUTION]: The skill manifest describes tools such as unbrowse.run and unbrowse.browse.open which provide broad browser automation and general task execution capabilities managed by the remote vendor service.

creativecommons