The Agent Skills Directory

[SAFE]: The skill provides a legitimate development utility for local task automation. Analysis of all files, including SKILL.md and the supporting documentation, confirms the absence of malicious code, data exfiltration patterns, or persistence mechanisms.
[COMMAND_EXECUTION]: The provided ralph.sh script executes claude-code (a tool from the trusted organization Anthropics) and standard project validation commands such as npm test and npm run typecheck. These executions are transparent and necessary for the functional requirements of an autonomous coding loop.
[PROMPT_INJECTION]: The agent instructions defined in prompt.md are focused on workflow management and implementation accuracy. No attempts to bypass AI safety guardrails, ignore system constraints, or extract system prompts were found.
[SAFE]: The skill defines a memory persistence strategy through local progress.txt and prd.json files. This approach is isolated to the project directory and does not access sensitive system files or environment variables.

ralph-loop-setup