eg
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructions include a method to install the
egCLI by piping a remote shell script directly from the author's GitHub repository into a shell (curl -fsSL ... | sh). - [COMMAND_EXECUTION]: The primary function of the skill is to execute arbitrary subprocesses using the
eg execcommand, which runs user-supplied commands with environment variables injected from specific profiles. - [EXTERNAL_DOWNLOADS]: The skill facilitates downloading software and configuration from the author's GitHub repository (
github.com/stjbrown/env-garden) and via Homebrew taps. - [PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it interpolates user-provided commands and profile names into shell executions.
- Ingestion points: User-provided
<profile>and<command>arguments inSKILL.md. - Boundary markers: The instructions recommend using the
--separator, but do not specify delimiters for the command content itself to prevent the agent from interpreting instructions within the data. - Capability inventory: Subprocess execution via
eg execinSKILL.md. - Sanitization: No specific sanitization or escaping rules are provided for the user-supplied command strings.
Audit Metadata