reap

What problem does it solve?

Reap turns completed engineering outcomes into structured, reviewable learning proposals, so teams can identify what should be added or updated in long-term knowledge without manually tracing raw evidence.

Core Features & Use Cases

• Post-epic learning extraction: Reads the build trinity after validate completes and proposes tiered learnings for human decision-making.
• Taxonomy-classified proposals: Produces proposals with a two-level learning taxonomy aligned to the knowledge base structure (learning_category + sub_category).
• Human-gated evidence commitment: Stages proposals in STM and requires a Tether/Vanish checkpoint before committing evidence/self-committing.
• Tier-aware outputs: Generates Tier 1 ADR draft drafts (with impact blocks) and supports tiered outcomes, including zero-proposal runs.

Quick Start

Run reap for a validated issue by invoking the command: /reap <issue>.