The Agent Skills Directory

[PROMPT_INJECTION]: The skill inherently processes external data (search results and web content) which is a vector for indirect prompt injection. If a search result contains instructions intended for an AI, the agent might inadvertently follow them. The author has included explicit safety notes in SKILL.md, EXAMPLES.md, and REFERENCE.md instructing the agent to treat results as untrusted remote input.
[COMMAND_EXECUTION]: The skill uses Bash as its execution environment. Documentation in EXAMPLES.md shows patterns where search results are piped into shell loops, using sed for filename generation and redirecting output to create local files. While these are common automation patterns, they represent a capability that could be abused if the agent is influenced by malicious search content.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to api.browserbase.com to perform searches and retrieve web pages. These are authenticated requests to the vendor's official API using the X-BB-API-Key header.

search