The Agent Skills Directory

REMOTE_CODE_EXECUTION (HIGH): The skill uses npx agentifind init, which downloads and executes code from the npm registry at runtime. Neither the author AvivK5498 nor the package agentifind belong to the predefined list of trusted sources.
COMMAND_EXECUTION (HIGH): The skill requires the agent to execute shell commands that perform heavy file system operations and configuration changes.
INDIRECT_PROMPT_INJECTION (HIGH): The skill instructs the agent to read 10-15 representative files from the repository to extract coding conventions. These files constitute untrusted data. Because the agent has already established command execution and file-writing capabilities in this context, malicious instructions embedded in the codebase (e.g., in code comments) could manipulate the agent into performing unauthorized actions.
CAPABILITY_INVENTORY (INFO): The skill requests permission to run npx (execution), read arbitrary project files (read), and write multiple configuration and documentation files to a .agent/ directory (write).

agentifind-init