The Agent Skills Directory

[PERSISTENCE]: The skill provides detailed instructions for modifying agent and IDE configuration files to ensure the Microlink MCP server is loaded in future sessions. Specifically, it guides the modification of claude_desktop_config.json, .vscode/mcp.json, and .cursor/mcp.json to add the @microlink/mcp server.
[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to fetch and process arbitrary content from external URLs (via methods like markdown, html, and text). This creates an attack surface where malicious instructions embedded in third-party websites could influence the agent's behavior.
Ingestion points: External content is ingested through microlink.markdown(), microlink.html(), and other scraping methods described in SKILL.md.
Boundary markers: The documentation does not specify the use of delimiters or warnings for the agent to ignore instructions within the fetched content.
Capability inventory: The agent can perform network requests, execute code via the function method, and modify local configuration files.
Sanitization: No explicit sanitization of fetched web content is described within the skill body.
[DYNAMIC_EXECUTION]: The skill documents the function (alias run) method and the evaluate extraction rule, which allow the agent to send and execute arbitrary JavaScript code within Microlink's remote browser sandbox.
[COMMAND_EXECUTION]: The documentation includes standard commands for installing and running the Microlink CLI and MCP server using npm install and npx, which are necessary for the skill's local execution capabilities.

microlink