The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the ralph-cli tool directly from a personal GitHub repository (https://github.com/mikeyobrien/ralph-orchestrator) using cargo install.
[COMMAND_EXECUTION]: The skill makes extensive use of local command execution, including tmux for session management, git for status monitoring, and the ralph CLI for running the orchestration loops.
[PROMPT_INJECTION]: The skill is designed to manage autonomous agents that communicate through shared markdown files, creating a surface for indirect prompt injection.
Ingestion points: Agents consume context from .kiro/specs/*/requirements.md, .ralph/agent/scratchpad.md, and .ralph/agent/memories.md.
Boundary markers: The instructions for the 'planner', 'builder', and 'reviewer' roles do not define clear boundary markers or explicit safety instructions to distinguish between trusted commands and data within the coordination files.
Capability inventory: The system has capabilities to execute shell commands, perform Git operations, and modify the project's file system through the ralph tool and the builder's instructions.
Sanitization: No sanitization or validation of the content within the requirements or scratchpad files is implemented before they are processed by the LLM agents.

ralph-loop