What problem does it solve? Memory files grow into oversized, flat documents that lose structure over time, making it hard to trace what was done, when, and with what proof. This Skill reorganizes them into a granular layout without losing identity, evidence, chronology, or critical decisions. ## Core Features & Use Cases - Convert mode: Moves flat memory files into memory/YYYY-MM-DD/ folders, splits oversized ledgers into one-thing-per-file entries, and indexes them under the ledger's ## Entries section. - Trim mode: Compacts ledgers over 1200 characters and entries over 800 characters by removing repeated plan text and duplicate evidence while preserving commits, PRs, and decisions. - Parallel wave delegation: Distributes conversion work across subagent workers with disjoint write sets, then validates deterministically by comparing commit hashes and field values before and after. - Use Case: During a weekly review, a ledger has grown to 2400 characters mixing delivery notes, telemetry, and decisions. The Skill splits it into a compact ledger plus four evidence-linked entries, verifies no commit hash was lost, and validates wikilinks. ## Quick Start Use optimize-memory to convert the oversized memory files in this project into the granular date-folder layout while preserving all evidence and decisions.
