The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of several Node.js packages (mcp-omnisearch, mcp-sqlite-tools, and mcpick). These resources are provided by the skill author and are legitimate tools for the described functionality.
[COMMAND_EXECUTION]: Provides shell commands for global package installation (npm install -g) and instructions to configure the agent to execute tools via npx. These are standard procedures for deploying CLI-based MCP servers.
[PROMPT_INJECTION]: The mcp-omnisearch tool introduces a surface for indirect prompt injection by fetching and processing content from the web.
Ingestion points: External data from search providers (Tavily, Brave, Kagi, Perplexity) and GitHub is ingested into the agent context via the mcp-omnisearch server.
Boundary markers: No explicit delimiters or boundary instructions are included in the configuration to isolate the untrusted search data.
Capability inventory: The agent is granted capabilities to perform SQLite operations (including write access) and manage the lifecycle of other MCP servers.
Sanitization: No sanitization, validation, or filtering mechanisms for the external search content are specified in the setup instructions.

mcp-setup