The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the installation of a global npm package @jackwener/opencli from an unverified third-party source. It also refers to a local developer path for a Chrome extension at ~/Desktop/code/AI/tools/opencli/extension/, which suggests a non-standard and unverified execution environment.
[COMMAND_EXECUTION]: User-provided input in the {{PROMPT}} placeholder is directly interpolated into shell commands (e.g., opencli grok ask "{{PROMPT}}"). This presents a command injection risk if the user input contains shell metacharacters or specifically crafted quotes designed to break out of the command string.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data entering the agent context via the output of the opencli command (Grok/Gemini responses). The analysis of the skill reveals:
Ingestion points: User input {{PROMPT}} and the resulting stdout from the opencli tool.
Boundary markers: Absent. The skill does not use delimiters or instructions to ignore embedded commands in the external AI responses.
Capability inventory: Uses the Bash tool to execute opencli and shell commands.
Sanitization: No escaping or validation is performed on the prompt or the retrieved response before processing.

ask-opencli