methodology-review

What problem does it solve?

It prevents an AI-assisted coding methodology from silently degrading as model capabilities change, by periodically validating which gates, rules, and workflow steps still meaningfully improve output quality.

Core Features & Use Cases

• Ablation-driven review: Tests whether removing or simplifying a component degrades correctness, completeness, and spec conformance, so you only cut what you’ve proven is redundant.
• Evidence-first verdicts: Produces a structured LOAD-BEARING / PARTIALLY REDUNDANT / NOT NEEDED / UNCERTAIN outcome to guide concrete updates and documentation.
• Learning loop integration: Logs findings to the repository learnings artifact and optionally submits methodology-applicable lessons to a Learnings Hub for reuse across projects.
• Scope-appropriate routing guidance: Helps decide whether new checks should live in rules, workflows, skills, or scripts based on how they’re invoked and enforced.

Quick Start

Run /methodology-review during a quarterly maintenance window to test one suspected component via the /experiment workflow and decide whether to keep, simplify, or remove it based on A/B evidence.