The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides documentation on how to manually install Homebrew using a piped execution pattern: /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)". While this fetches the official installation script from Homebrew's GitHub repository, it constitutes a remote code execution pattern.
[EXTERNAL_DOWNLOADS]: The workspace_download tool enables the agent to fetch binary content, such as PDFs and images, from external URLs and save them to the local workspace.
[COMMAND_EXECUTION]: The skill defines a safety framework for CLI execution, identifying tools like pip install, brew install, curl, and wget as requiring an explicit approval_card from the user.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks due to its data processing capabilities:
Ingestion points: External data enters the context through workspace_read (reading workspace files) and workspace_download (fetching content from URLs).
Boundary markers: The instructions do not specify the use of delimiters or instructions to ignore embedded commands within ingested data.
Capability inventory: The agent has the ability to execute shell commands (via bash) and perform file system operations (workspace_save, workspace_append).
Sanitization: There are no explicit requirements for sanitizing or validating the content retrieved from external files or URLs.

Workspace SOP