What problem does it solve? When an AI agent states a wrong fact, simply noting the correction leaves the bad data in place to resurface later. This Skill traces every user-reported factual error back to its true source — a brain page, memory file, rendered identity file, or facts table — fixes it there, and checks for propagation. ## Core Features & Use Cases - Root-Cause Classification: Sorts each error into BRAIN_ERROR, MEMORY_ERROR, SOUL_USER_ERROR, FACTS_TABLE_ERROR, HALLUCINATION, STALE_DATA, or CROSS_CONTAMINATION, each with its own fix surface. - Source-Level Fixes: Edits brain pages with commit and re-sync, expires wrong facts rows via forget/remember with provenance, and fixes rendered SOUL.md/USER.md through the bootstrap answer bank rather than direct edits. - Propagation Checks: Greps the whole brain and runs gbrain search to find and fix every instance of the wrong fact, then delivers a structured root-cause report. - Use Case: A user says "That's wrong — Alice never worked at WidgetCo." The Skill searches the brain and memory, finds the bad claim on a people page, corrects it with a dated source note, re-syncs, checks for spread, and reports the fix. ## Quick Start Tell the agent "that's wrong, root-cause this error and fix it everywhere" after it states an incorrect fact.
