Auto-Memory Review and Cleanup Plan
Get an evidence-backed plan to promote durable memories, merge duplicates, flag conflicts, and prune stale entries.
Before & after
Same prompt, same agent — the only difference is the skill.
That's a thoughtful question. You'd need to look back at what you've written down or what Claude has learned about you over time. Some things are important enough to remember long-term, others are just temporary notes that you can delete. You might have some insights that keep coming up and should be promoted to a more persistent place. There could be duplicate entries that should be consolidated. It's worth doing this review periodically to keep your memory clean and useful. Have you written down any learnings recently? What areas are most important to remember? Do you track your goals or habits anywhere? I can help you think through what matters most.
AUTO-MEMORY ANALYSIS TOTAL ENTRIES: 124 active records Last review: 45 days ago HEALTH METRICS: Age distribution: 18 recent (0-30 days), 42 active (1-6 months), 64 stale (6+ months) Redundancy: 7 duplicate/near-duplicate entries found Priority balance: 34 high, 56 medium, 34 low Coverage gaps: 4 areas needing attention PROMOTION CANDIDATES (should be permanent): 1. Decision framework for contract negotiations (3 references, last used 12 days ago) 2. Company communication style (referenced in 8 outputs) 3. Revenue calculation formula (used monthly) 4. Customer value prop positioning (referenced in 6 pieces) RECOMMENDATION: Promote these 4 to core knowledge STALE ENTRIES (consider pruning): 1. Q1 hiring timeline (outdated, 4 months old) 2. Pricing experiment notes (superseded by new data) 3. Vendor comparison from March (replaced by August vendor) 4. Three old meeting notes without follow-up RECOMMENDATION: Archive these 4, delete meeting notes CONSOLIDATION OPPORTUNITIES: Found 7 partial duplicates about customer feedback Merge into single structured entry Result: 3 entries saved, cleaner history COVERAGE GAPS: Missing: 2026 goals and milestones Missing: Updated competitive landscape notes Missing: Current product roadmap reasoning RECOMMENDATION: Add 3 new entries this week ACTION PLAN: ✓ Promote 4 entries ✓ Archive/delete 7 stale items ✓ Consolidate 7 duplicates ✓ Add 3 new gap entries Post-review entry count: 113 (more focused, more usable)
About this skill
Your agent keeps remembering an old decision, repeating the same fact, or treating a temporary workaround as a permanent rule. This skill audits the actual memory entries and produces a concrete cleanup plan, with replacement wording and evidence for each recommendation.
What it does
- Separates superseded facts from old entries that still apply, using dates, scope, and source authority.
- Finds semantic duplicates and drafts one canonical entry without dropping exceptions or conditions.
- Identifies durable practices worth promoting to project guidance and specifies their intended scope.
- Flags contradictions, missing evidence, sensitive content, and temporary states needing revalidation.
- Calculates review coverage, removable duplicate rate, confirmed stale rate, and unresolved entries with explicit denominators.
- Produces a reviewable action list and checks the result after any authorized edits.
What you get
A concise, evidence-backed plan grouped into promote, merge, prune, keep, and verify. Merge and promotion recommendations include proposed wording; deletion candidates remain pending confirmation.
Best for
Developers and teams using Claude Code or Claude whose accumulated memory contains conflicting decisions, duplicated notes, temporary workarounds, or practices ready for durable project guidance.
SKILL.md preview
---
name: auto-memory-review-and-cleanup-plan
description: >-
Audit agent auto-memory using claim-level evidence, scope-aware conflict checks, semantic deduplication, and promotion criteria; return exact replacement text and a reviewable cleanup plan. Use when the user says "review my memory", "what have you learned?", "my agent keeps remembering old decisions", "clean up duplicate memories", "what should move into CLAUDE.md?", "which memories are stale?", or "/si:memory-review".
---
# Auto-Memory Review and Cleanup Plan
… (sign up to view the full skill)More development skills
View all Development skills →Performance Bottleneck Profiling: Node, Python, Go
Classifies a slowdown, gives exact profiler commands, reads the output, and proves the fix with same-conditions p95/p99 numbers.
Feature Scaffolding with Working Tests
Turn a short feature spec into code, runnable tests, and integration wiring that match your existing project.
Feature Flag Rollouts and Kill Switches
Design staged flag rollouts, verify rollback paths, and audit flag debt with explicit defaults, gates, and retirement plans.
Microcontroller Firmware and RTOS Engineering
Build and debug MCU firmware with bounded ISR/RTOS timing, safe DMA ownership, and measured power budgets.