From claude-memory
This skill should be used when the user asks to consolidate memories, merge observations, synthesize insights, run consolidation, or clean up memory. Trigger phrases include "consolidate memories", "merge observations", "run consolidation", "synthesize insights", "clean up memory".
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-memory:memory-consolidateThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Manually trigger memory consolidation to synthesize observations into insights.
Manually trigger memory consolidation to synthesize observations into insights.
Before consolidating, check how many unconsolidated observations exist. Consolidation requires at least 3 unconsolidated observations.
If the user wants to preview, or if {{dry_run}} is "true", run a dry run first:
~/.config/claude-memory/.venv/bin/python3 ${CLAUDE_PLUGIN_ROOT}/scripts/consolidate.py --dry-run 2>/dev/null || python3 ${CLAUDE_PLUGIN_ROOT}/scripts/consolidate.py --dry-run
This outputs JSON showing:
unconsolidated_count — how many observations are pendingthreshold — minimum needed (3)would_consolidate — whether consolidation would proceedobservations — list of observations that would be consolidated (id, content preview, priority, topics, importance)Present this as a readable preview to the user. If would_consolidate is false, explain that more observations are needed.
If the user confirms (or didn't request dry run), execute consolidation in the foreground:
~/.config/claude-memory/.venv/bin/python3 ${CLAUDE_PLUGIN_ROOT}/scripts/consolidate.py --foreground 2>/dev/null || python3 ${CLAUDE_PLUGIN_ROOT}/scripts/consolidate.py --foreground
This outputs JSON with the consolidation result:
summary — Synthesized summary across related observationsinsight — Key pattern or insight discoveredconnections — Links found between observationssource_ids — IDs of observations that were consolidatedredundant_ids — Observations marked as redundant (importance reduced)observation_count — Total observations that were processedIf the result is {"status": "no_consolidation", ...}, explain that consolidation couldn't proceed (not enough observations, or LLM call failed).
Show the user:
GOOGLE_API_KEY, ANTHROPIC_API_KEY, etc.)npx claudepluginhub seangsisg/deja-claudeReviews conversation, extracts learnings, and persists to auto-memory or vault. Run after substantial work sessions to avoid knowledge rot.
Audits Pensyve memories for staleness (>30 days unaccessed or low retrievability), contradictions, low confidence (<0.5), and consolidation candidates, then offers confirmed cleanup actions. Use periodically.
Optimizes neural memory through usage pattern analysis, recall performance, bottleneck detection, and consolidation/pruning/enrichment actions tracked via checkpoint Q&A.