experiment-iterative-coder

What problem does it solve?

Automates disciplined, repeatable code improvement cycles by orchestrating plan → code → evaluate → refine workflows that embed linting, testing, and self-evaluation to raise code quality in complex tasks.

Core Features & Use Cases

• Structured iteration: Decomposes tasks into sequential phases with up to 3 iterations per phase (max 10 iterations overall) to progressively elevate correctness and robustness.
• Automated quality gates: Runs lint checks (ruff), executes tests (pytest), and performs self-evaluation to surface targeted improvements.
• Decision-driven planning: Diagnoses failures and crafts precise improvement plans, enabling rapid, data-guided code refinement across multi-file projects.
• When to use: Ideal for high-stakes coding tasks, refactoring efforts, or scenarios where incremental quality gains are prioritized over single-pass output.

Quick Start

Tell the agent to start a code task with MODE: MORE_EFFORT and follow the plan → code → evaluate → refine cycle until quality targets are met.