The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs network requests to the Brave Search API and arbitrary external URLs to retrieve search results and page content. This behavior is essential to its intended purpose and does not involve unauthorized access to sensitive local data.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it processes and presents untrusted content from the web to the agent.
Ingestion points: URL fetching in content.js and the fetchPageContent function in search.js.
Boundary markers: Absent; the skill does not use specific delimiters or instructions to ignore commands within the extracted web content.
Capability inventory: Code analysis across search.js and content.js shows no use of subprocesses, eval(), exec(), or file writing operations.
Sanitization: Present; the skill utilizes jsdom and @mozilla/readability to extract core article content and strip potentially malicious HTML elements such as scripts and styles, which significantly mitigates basic injection risks.

brave-search