memory-audit

Audit EDITH memory layers and flag stale or contradictory stored facts.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/knsiuss/orion --skill memory-audit-knsiuss
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: memory-audit
Source: https://github.com/knsiuss/orion/tree/main/workspace/skills/memory-audit
Command: npx skills add https://github.com/knsiuss/orion --skill memory-audit-knsiuss

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you review, manage, and clean up the persistent memory EDITH uses to store information about you and your interactions, ensuring privacy and accuracy.

Core Features & Use Cases

  • Memory Review: Inspect all stored facts across different memory layers (Pinned Facts, User Profile, Vector Store, Episodic Logs, Prisma DB).
  • Stale Entry Detection: Automatically flags facts older than 90 days for review.
  • Contradiction Detection: Identifies conflicting information stored in different memory layers.
  • Cleanup Interface: Provides an interactive way to delete or correct outdated or incorrect information after confirmation.
  • Use Case: You want to ensure EDITH doesn't hold onto outdated personal information. You can use this skill to see everything EDITH remembers, identify anything that's no longer relevant, and remove it.

Quick Start

Ask EDITH to audit your memory and flag anything outdated.

Frequently Asked Questions about memory-audit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit and clean up persistent memory facts stored across different layers?

To audit memory facts, this Skill reviews stored data across Pinned Facts, User Profile, Vector Store, Episodic Logs, and Prisma DB to detect stale entries and flag contradictions for deletion or correction.

How do I detect and resolve contradictions in stored user profile memory?

Contradiction detection scans all memory layers to identify conflicting information stored across different sources. It presents the conflicting facts so you can review and correct the outdated or incorrect data.

What is the best way to remove outdated personal information from persistent memory?

The best way to remove outdated personal information is using the cleanup interface, which automatically flags facts older than 90 days for review and provides an interactive method to delete them after confirmation.

Can I review episodic logs and vector store data to ensure memory privacy and accuracy?

Yes, memory review inspects stored facts across all layers including Episodic Logs and Vector Store. You can see everything stored, identify irrelevant entries, and remove them to ensure data privacy and accuracy.

What happens to stale memory entries older than 90 days during a memory audit?

Stale memory entries older than 90 days are automatically flagged for review during the audit. The interface then presents these flagged items, allowing you to confirm deletion or apply corrections to the outdated facts.

Does memory cleanup require manual confirmation before deleting facts from the Prisma DB?

Yes, the cleanup interface requires manual confirmation before deleting or correcting any facts. This ensures you maintain control over data removal across all layers including Prisma DB, preventing accidental data loss.