The Agent Skills Directory

[SAFE]: The skill manages local project state using standard file operations on paths like .ralph-state.json and specs/. These actions are consistent with the skill's primary purpose.
[SAFE]: The skill includes a non-negotiable requirement to stop execution and wait for user approval after the setup phase, preventing unauthorized autonomous activity.
[SAFE]: All external inputs are limited to local configuration files and command-line flags, with no evidence of remote code execution or data exfiltration.

ralph-specum-start