Full-weight anti-loop protocol for complex, multi-step tasks. Eliminates procrastination, false completions, and repetitive loops. The agent executes, verifies, updates — without the user having to ask. For simple 1-3 step tasks, use ralph-small instead.
Mid-weight anti-loop protocol for moderate tasks (4-10 steps). Tracks attempts, forces method diversity, provides progress updates. Escalates to ralph-huge after 3 failed attempts. For 1-3 step tasks use ralph-small. For 10+ step / high-complexity tasks use ralph-huge.
Entry point for the Ralph anti-loop system. The agent assesses the incoming task and loads the correct anti-loop skill level automatically. Invoke at the START of every task, before any other action.
Lightweight anti-loop discipline for simple/routine tasks. Prevents stalling, false completions, and wasted tokens on quick operations. For complex multi-step tasks, escalate to the full ralph-huge skill.
Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) The user corrects you ('No, that's wrong...', 'Actually...'), (3) The user requests a capability that doesn't exist, (4) An external API or tool fails, (5) You realize your knowledge was outdated or incorrect, (6) A better approach is discovered for a recurring task. Also review learnings before major tasks.