What problem does it solve? AI coding clients accumulate stale, duplicated, or private memories that keep influencing future sessions. This Skill audits those memory stores, classifies what is worth keeping, and retires obsolete guidance only through each client's supported update mechanism. ## Core Features & Use Cases - Scoped memory inventory: Discovers configured Codex and Codex Lab stores, deduplicates aliases, and records ownership, consumed layers, and update contracts without disk sweeps. - Classification and promotion: Routes candidates to skills, repo docs or GitHub issues, local config, or retirement, verifying each claim against maintained sources first. - Staged retirement evidence: Distinguishes requested, recorded, applied/regenerated, and retrieval-verified stages, requiring a fresh session to prove stale guidance no longer governs answers. - Use Case: A user notices Codex keeps recommending an old deployment process. They invoke this Skill to audit the memory store, propose retiring the stale claim, record an approved retirement note, and verify in a fresh session that the corrected guidance is retrieved instead. ## Quick Start Ask the agent to audit my Codex memories and propose which entries should be promoted to skills or retired from reusable context.
