The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill utilize a shell script from https://install.meteor.com/ which is piped to sh for installation on Linux and macOS hosts. As this script originates from the official vendor domain for the authoring organization (meteor), it is evaluated as a trusted deployment resource and does not escalate the severity of the verdict.
[PRIVILEGE_ESCALATION]: The instructions permit the use of sudo for environment-specific repair tasks, particularly when global npm permissions prevent standard installation. This capability is strictly gated by instructions requiring explicit user authorization and acknowledgment of root-level risks.
[COMMAND_EXECUTION]: The skill executes multiple system-level commands through the meteor CLI and npx, including meteor reset and npx meteor uninstall. Security is managed through a mandatory decision flow that requires human-in-the-loop confirmation before any operation that removes project or user state.
[INDIRECT_PROMPT_INJECTION]: The skill consumes untrusted data from user prompts regarding environment setup and reads content from project configuration files (.meteor/release). 1. Ingestion points: User set-up descriptions and project-local release files. 2. Boundary markers: Present; the skill defines a rigid decision flow to classify requests before action. 3. Capability inventory: System command execution for tool installation and environment modification (PATH). 4. Sanitization: Absent; the skill relies primarily on user confirmation for any non-trivial or destructive system changes.

meteor-cli-installation