The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in hooks/thread.py to interact with system clipboard utilities such as pbcopy, xclip, xsel, wl-copy, or PowerShell. This is a legitimate and necessary use of command execution to implement the skill's core functionality of placing text onto the user's clipboard.
[INDIRECT_PROMPT_INJECTION]: The skill processes conversation logs (.jsonl files) which may contain untrusted data from previous messages. It formats this data into a transcript for the user. While this creates a theoretical attack surface, the risk is mitigated because the output is specifically intended to be wrapped in a code block for the user to copy or sent directly to the clipboard, rather than being parsed as instructions for the agent's next action.
[SAFE]: The skill accesses local application configuration and log files located in the ~/.claude/ directory. This data access is essential for generating the transcript and determining the user's name for formatting purposes. The implementation is privacy-conscious, only extracting the display name and conversation text without any network exfiltration.

copy-chat