The Agent Skills Directory

[SAFE]: The skill's primary purpose is to scaffold a local automation harness. It copies templates and scripts from the plugin directory to scripts/ralph/ in the current repository.- [COMMAND_EXECUTION]: The skill utilizes a bash script (ralph.sh) to manage the autonomous loop. This script interacts with the claude CLI, git, and local files to coordinate tasks. This is standard behavior for a CLI-based automation tool.- [DATA_EXFILTRATION]: No exfiltration detected. The script identifies the user and host for local logging purposes but does not transmit sensitive data to external servers.- [EXTERNAL_DOWNLOADS]: The skill does not perform any network downloads. It relies on pre-installed tools like python3, git, and the claude CLI.- [PROMPT_INJECTION]: Not detected. The instructions in the prompt templates and the ralph.sh script are designed to guide the sub-agent's behavior within the autonomous loop, not to bypass safety guidelines.

flow-next-ralph-init