The Agent Skills Directory

[COMMAND_EXECUTION]: The skill dynamically constructs and executes shell commands to start web servers based on the detected project type, such as npm run dev, php -S localhost:8000, python manage.py runserver, and cargo run.\n- [REMOTE_CODE_EXECUTION]: The skill explicitly instructs the agent to search for and execute startup commands found within documentation files such as README.md, AGENTS.md, and .monkeycode/MEMORY.md. If these files originate from an untrusted source, such as a malicious or compromised repository, they could contain destructive shell commands that the agent will execute in the user's local environment.\n- [PROMPT_INJECTION]: Indirect prompt injection vulnerability. The skill ingests instructions from untrusted workspace files to determine its execution logic without sanitization or strict boundary markers. Ingestion points include README.md, AGENTS.md, and .monkeycode/MEMORY.md. The skill has access to the background_terminal_create tool for arbitrary shell execution and does not perform validation against a whitelist of safe commands before extracting and executing strings from the documentation.\n- [EXTERNAL_DOWNLOADS]: The skill executes package managers including npm, pip, composer, and bundle to download and install dependencies from external registries based on the configuration files found in the workspace (e.g., package.json, requirements.txt).

deploy-website