The Agent Skills Directory

[COMMAND_EXECUTION]: User input from the [query] argument is interpolated directly into shell commands within the SKILL.md instructions, such as find, grep, and git log. If the query contains shell metacharacters, it could lead to arbitrary command execution on the host system.\n- [DATA_EXFILTRATION]: The scripts/dig.py script reads Claude Code session history from .jsonl files. These files contain full logs of previous interactions, which may include sensitive information or code. Accessing these files exposes private historical data to the agent's current context.\n- [EXTERNAL_DOWNLOADS]: The skill supports cloning external repositories using ghq get when a URL is provided to the --repo argument. This allows for the retrieval and local storage of external code from arbitrary sources.

trace