The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard macOS command-line tools to interact with the system. It uses pbpaste to retrieve content from the clipboard and pbcopy to return the cleaned version. It also uses the open command to launch generated HTML files in the user's default web browser.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted data from the system clipboard.
Ingestion points: The pbpaste command reads arbitrary text from the clipboard into the agent's context (SKILL.md, Step 1).
Boundary markers: The instructions do not define specific delimiters to isolate clipboard content from the agent's instructions.
Capability inventory: The skill can execute shell commands (pbcopy, open) and write files to the /tmp/ directory.
Sanitization: The skill implements significant sanitization by stripping ANSI escape codes and terminal artifacts. Furthermore, it includes a dedicated secret scanning step that uses regex patterns to detect and mask API keys or tokens before they are processed or copied.
[DYNAMIC_EXECUTION]: For rich-text destinations like email or HTML, the skill generates temporary HTML files in the /tmp/ directory. It may also leverage pandoc, if installed on the system, to perform markdown-to-HTML conversion.

paste