The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes pnpx to download and run the @sveltejs/mcp package at runtime. This package is the official Model Context Protocol implementation provided by the Svelte organization.
[COMMAND_EXECUTION]: Accesses system capabilities by executing CLI tools for listing documentation sections, retrieving content, and running code analysis. The instructions specifically include guidance on escaping shell characters (e.g., escaping $ as \$) to prevent accidental command injection or shell variable substitution when passing code to the terminal.
[INDIRECT_PROMPT_INJECTION]: The skill provides the svelte-autofixer tool which processes external Svelte code or files. This creates a surface where malicious instructions could be embedded within component code to influence the agent's behavior during analysis.
Ingestion points: Code snippets or file paths passed as arguments to pnpx @sveltejs/mcp svelte-autofixer in SKILL.md.
Boundary markers: Absent; the content is passed directly as a string argument within the shell command.
Capability inventory: The skill can execute Node.js tools that read local files and generate code analysis output for the agent to process.
Sanitization: While shell-level character escaping is advised for safety, the skill does not implement logical sanitization or filtering of the Svelte code content to detect or neutralize embedded prompt injection attempts.

svelte-code-writer