Claude Memory Framework — Development skill for Claude Code

claude-memory-framework

**A persistent memory layer for Claude Code that survives across sessions.**

Claude Code is powerful, but it forgets everything between sessions and loses context in long ones. This framework fixes both problems with a structured system of documents that Claude reads on startup and updates before closing.

Session 1                    Session 2                    Session 3
─────────                    ─────────                    ─────────
Read docs → Work →           Read docs → Work →           Read docs → Work →
Update docs                  Update docs                  Update docs

**Architected by [Eugenio Ruiz de la Cuesta](https://www.linkedin.com/in/eugenio-ruiz-de-la-cuesta-fdez/) (orchestrating Claude Opus 4.6)** · v0.2.3 · [Unlicense](LICENSE) [@guachoxx](https://github.com/guachoxx)

---

Why

Problem	Without the framework	With the framework
New session	Claude starts from zero	Claude reads Project Index, picks up where you left off
Long session	Context fills up, hallucinations start	Distillation protocol saves progress before context runs out
Technical decisions	Lost in conversation transcripts	Persisted in structured documents
Onboarding Claude to your code	Re-explain every time	Module context documents provide instant context

Key Features

• 3-layer memory architecture — Root index → reference documents → per-module context. Claude loads only what it needs.
• Session distillation protocol — Systematic process to save progress before context runs out. Triggered automatically or on demand.
• 7 project documents — Specifications, Analysis, Plan, Technical Report, Testing, Status, Changelog — each with a clear audience and lifecycle.
• Provider-agnostic — Documents can live as local .md files, in ClickUp, Notion, or any other platform. The methodology stays the same.
• Lite mode — Small projects use