The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the execution of the arks CLI tool via the system shell to perform all web operations. Evidence: Multiple instances of arks invoke commands are used across references/code-context.md, references/extract.md, references/fetch.md, references/search.md, and references/site.md.
[EXTERNAL_DOWNLOADS]: The skill interacts with external web services and retrieves content from arbitrary URLs provided by the user or discovered via search. Evidence: The skill uses providers like Exa, Tavily, and Firecrawl to perform searches and crawl websites as described in SKILL.md and references/site.md. The web.fetch capability in references/fetch.md allows retrieving content from external HTTP(S) URLs.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the public web, creating a surface for indirect prompt injection attacks. * Ingestion points: Content retrieved via web.fetch (references/fetch.md), web.search/web.related (references/search.md), web.crawl/web.map (references/site.md), and web.extract (references/extract.md). * Boundary markers: Absent; the skill instructions do not define specific delimiters or instructions for the agent to treat the retrieved content as untrusted data separate from its own instructions. * Capability inventory: The agent has the ability to execute local shell commands using arks invoke. * Sanitization: Absent; the skill lacks instructions for sanitizing or filtering the fetched content before the agent processes it.

web