Analyse a failure, capture the debugging journey, and store the lesson as a persistent memory. Use after low L-Scores, multi-round debugging, or any significant mistake.
Analyse what went wrong, capture the debugging journey, and store the lesson so it's never repeated.
Requires a description of what failed:
/post-mortem the FalkorDBLite create_relationship was broken due to Cypher preamble format
/post-mortem dashboard pipeline tab showing 0/50 agents despite active pipeline
If no argument is provided, ask: "What failure should I analyse?"
NOTE: This analysis uses what's visible in the current conversation context. If the session was compacted or the failure happened in a previous session, provide the description explicitly — do not rely on context inference.
Work through these in order:
What was the observable problem? What did the user see or report?
What was tried first? Why did it fail? What was tried next? Capture each attempt:
What was the actual underlying problem? Not the symptom — the cause.
What ultimately resolved it?
The non-obvious lesson. The thing that would have saved time if known upfront. This is the most valuable part.
Call mcp__memorygraph__store_memory with:
fixPost-mortem: [brief description]["post-mortem", "debug-journey", "[relevant tech tags]"]Then search MemoryGraph for related memories and create RELATED_TO relationships where relevant.
Post-mortem stored: [memory_id]
Symptom: [one line]
Root cause: [one line]
Key insight: [one line]
Linked to: [N] related memories