Project Memory Template — Development skill for Claude Code

project-memory-template

A reusable, complexity-scaled project-memory scaffold — `CLAUDE.md` + `PLAN.md` + `.claude/memory/` + `CONTEXT.md` + `docs/adr/`, sized to how much a project actually needs, not applied uniformly.

Synthesized from `Python-PowerBI-DynastyFantasyFootball`'s real memory architecture (built organically over months, growing a piece at a time as each need showed up) and validated against `skills-plugins-hooks-agents`' vendored engineering flow (`grill-with-docs`/`domain-modeling`, which generate `CONTEXT.md`/ADR content live rather than pre-scaffolding it blank) and ponytail's YAGNI ladder (introduce complexity only as it's earned).

Dependency

This template assumes the central skills library, [`skills-plugins-hooks-agents`](https://github.com/benjamininja/skills-plugins-hooks-agents) (`ask-matt`, `grill-with-docs`, `domain-modeling`, `tdd`, etc.), is installed — see that repo's README for the symlink-installation steps. Every tier's `CLAUDE.md` names this dependency explicitly so a new repo built from this template doesn't silently assume skills exist without saying where they come from.

The three tiers

See [`tiers/`](tiers/) for the actual files, and [`docs/graduating-tiers.md`](docs/graduating-tiers.md) for exactly when to add each piece — the additions are **independent triggers**, not a strict ladder.

• tiers/minimal/ — CLAUDE.md + PLAN.md. A small script or short-lived tool with no real data model or terminology to track.
• tiers/standard/ — adds .claude/memory/MEMORY.md (project state index) + CONTEXT.md (domain glossary). Ongoing state worth remembering, but not yet enough decision-density for formal ADRs.
• tiers/full/ — adds docs/adr/ (numbered decision records) + the token-gating execution-loop section in CLAUDE.md. Real architectural decisions with alternatives worth defending later.

How to use

Copy the contents of the tier that fits your project into its repo root, fill in the