nhadaututtheky/neural-memory
3 skills1.8k installs609 starsGitHub
Install
npx skills add https://github.com/nhadaututtheky/neural-memorySkills in this repo
1Memory Evolutionmemory-evolution is an agent skill for builders using NeuralMemory who want memories to stay useful as projects evolve. It instructs the agent to act as a Memory Evolution Specialist: collect evidence from recall tooling, classify which memories are hot, cold, or dead, diagnose bottlenecks that confuse retrieval, and propose specific graph operations such as consolidation, pruning, and enrichment. When you pass no focus argument, it runs the full evolution cycle end to end. Outputs are structured as usage analysis, bottleneck report, evolution actions, and a checkpoint log for the next tuning pass. That matters for solo builders relying on persistent agent context—you avoid bloated or contradictory memory that slows every coding session.651installs2Memory IntakeMemory Intake is a NeuralMemory specialist workflow that converts messy notes, conversations, and unstructured thoughts into well-typed, tagged, confidence-scored memories. Solo builders running Claude Code with a ~/.neuralmemory setup use it when raw capture would pollute recall—instead of one giant paste, the skill classifies each information unit, asks clarifying questions one at a time, and stores a coherent batch through nmem_remember. Outputs include a categorized intake report, identified ambiguities, and connections to memories already on file. It fits anywhere in the solo-builder journey you need durable agent memory: competitor snippets in Idea, scope decisions in Validate, stack choices in Build, launch checklists in Ship, or ops runbooks in Operate. The pattern is librarian-grade filing so nmem_recall returns the right fact at the right moment without manual re-explaining.650installs3Memory AuditMemory-audit is an agent skill package for builders who rely on NeuralMemory as durable agent context. It instructs the model to act as a Memory Quality Auditor: collect baseline metrics with `nmem_stats` and `nmem_health`, optionally deepen with recall, context, and conflict tools, then score the brain across six dimensions and emit a graded report. Findings are prioritized by severity with concrete memory references so you can delete, merge, or rewrite neurons instead of guessing. Recommendations are ordered by impact and may include projected metric shifts if you apply them. Run it when responses feel stale, contradictory, or overly broad—or on a schedule as part of agent ops. It assumes NeuralMemory is installed and tool access is configured; it is not a substitute for backup policy or privacy review of what you store in memory.519installs