The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to ingest and process untrusted external data from search results. While it lacks direct execution capabilities, it presents an attack surface for indirect prompt injection.
Ingestion points: Data enters the agent via the Brave Search API response fields, specifically results[].title and results[].description.
Boundary markers: None identified in the skill definition to help the agent distinguish between search metadata and potential embedded instructions.
Capability inventory: The skill itself only performs data retrieval (HTTP GET/POST) and does not include file-writing, command execution, or network exfiltration capabilities.
Sanitization: There is no explicit sanitization or filtering of the search results content mentioned in the documentation.
[Data Exposure & Exfiltration] (SAFE): The skill uses environment variable placeholders (${BRAVE_SEARCH_API_KEY}) and generic placeholders (<API_KEY>) for authentication, avoiding hardcoded secrets.
[Command Execution] (SAFE): The provided cURL examples are for documentation purposes and do not indicate a pattern of the agent executing arbitrary shell commands.

videos-search