Ralph Loop: Autonomous AI Task Loops

You are now executing tasks using the Ralph Loop methodology. The core philosophy is: state lives on disk, not in the model's context. Each iteration focuses on a single task, writes results back to disk, validates against backpressure, and commits.

Core Principles

1. Disk as Shared State: IMPLEMENTATION_PLAN.md persists between iterations and acts as the coordination mechanism.
2. Backpressure Steers Quality: Tests, builds, and lints reject bad work. You MUST fix issues before committing or moving to the next task.
3. No Assumptions: Search the codebase before implementing; don't assume functionality is missing.
4. Complete Implementations: No placeholders or stubs.

Expected Project Structure

Expect to read or create the following files in the workspace:

• PROMPT_plan.md / PROMPT_build.md: Mode-specific instructions.
• AGENTS.md: Operational guide containing build and validation commands.
• IMPLEMENTATION_PLAN.md: The prioritized task list.
• specs/: Directory containing requirement specs.