The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill recommends executing an unverified Node.js package submit-expo-feedback via npx --yes. This package is not a recognized tool from the official Expo organization nor does it follow the author's identified naming patterns, creating a potential vector for executing arbitrary code from an untrusted source.
[DATA_EXFILTRATION]: The instructions for submitting feedback suggest including 'actionable feedback' and 'relevant context'. An AI agent following these instructions might transmit sensitive information such as source code snippets, environment variables, or private configuration data to an external server controlled by the package author.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands to scaffold modules and interact with the system. It directly interpolates user-provided strings into these commands, which could lead to command injection if the input is not carefully handled.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted user data to generate and execute code.
Ingestion points: User instructions for module names, features, and implementation details in SKILL.md and create-expo-module.md.
Boundary markers: None; there are no instructions to the agent to treat user data as untrusted or to delimit external input.
Capability inventory: File-write capabilities (scaffolding), network operations via npm/npx, and subprocess execution.
Sanitization: The skill does not provide methods for sanitizing or escaping user-provided content before it is used in shell commands or native code files.

expo-module