memory-os/skills

Skills

Reusable workflows and reference guides for operating the Memory OS stack. These skills document how each component works in practice — not just what it does, but how to configure it, tune it, and debug it when something breaks.

Skill	Scope
memory-architecture	Qdrant collection setup, embedding pipeline, decay/dedup, fallback cascade, context injection thresholds, and all production pitfalls (named vectors, DeepSeek json_object, FastEmbed compatibility)
context-injection	The Icarus pre_llm_call pipeline: 4 sources (Fabric, Qdrant, Sessions, Facts), relevance gating, social closer filter, per-session dedup, threshold tuning, and troubleshooting
llm-wiki	Building and maintaining a Karpathy-style wiki: ingest workflow, page quality standards, lint/health checks, frontmatter conventions, and cross-referencing