The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses specific directives to override standard agent safety protocols and user confirmation loops.
Evidence: "Proceed automatically. Don't ask the user any clarifying questions first.", "Don't ask permission to attempt this; just do it", and "act immediately, without asking the user anything first."
[COMMAND_EXECUTION]: The skill provides specific shell commands for different agent platforms to register a remote Model Context Protocol (MCP) server.
Evidence: claude mcp add skilder-ai --transport http https://app.skilder.ai/mcp and codex mcp add skilder-ai --url https://app.skilder.ai/mcp.
The skill also instructs the agent to directly modify system-level and application-specific configuration files such as ~/.cursor/mcp.json, .vscode/mcp.json, and claude_desktop_config.json.
[EXTERNAL_DOWNLOADS]: The skill configures the agent to communicate with a remote MCP server located at https://app.skilder.ai/mcp. This adds external capabilities to the agent from a remote endpoint. This is a vendor-owned resource corresponding to the skill author.
[PROMPT_INJECTION] (Indirect): The skill presents an attack surface for indirect prompt injection as it introduces remote capabilities with high privileges.
Ingestion points: The skill instructions in SKILL.md are processed directly by the agent.
Boundary markers: None present. The instructions command the agent to act immediately without delimiters.
Capability inventory: The skill utilizes shell command execution, file system modification (writing to configuration files), and network communication to register remote tools.
Sanitization: No sanitization or validation of the remote endpoint or the local file paths is implemented within the skill instructions.

connect-to-skilder