Context Packs

Domain-specific knowledge packs for Claude Code. Load relevant packs based on your session topic.

Usage

In semianalysis-claude

Context packs are loaded automatically based on your session goal. The shared pack loads every session; domain packs load as needed.

Standalone

# Clone to your project
git clone https://github.com/Semianalysis/Context-Packs.git context-packs

# Reference in your CLAUDE.md
# "Load context from context-packs/shared/ and context-packs/memory/"

Available Packs

Pack	Description	Load When
shared	Style guide, terminology, data sources	Always
semiconductors	Industry overview, supply chain, process nodes	Broad semi coverage
memory	DRAM, NAND, HBM deep dive	Memory-specific analysis
ai-accelerators	GPUs, TPUs, custom silicon	AI hardware coverage
foundry	TSMC, Samsung, Intel, packaging	Manufacturing analysis
datacenter	Infrastructure, hyperscalers, power	DC buildout coverage
equipment	ASML, Applied, Lam, KLA	Equipment/litho analysis
geopolitics	Export controls, CHIPS Act, China	Policy/trade analysis
networking	NVLink, InfiniBand, CPO, interconnects	AI cluster networking
hyperscalers	Google, Microsoft, Amazon, Meta strategies	Cloud/custom silicon
finance	Financial statements, valuation, metrics	Company analysis

Pack Structure

Each pack contains:

{pack}/
├── CONTEXT.md      # Overview, when to load, key concepts
├── players.md      # Key companies, relationships (optional)
├── terminology.md  # Domain-specific terms (optional)
└── sources.md      # Where to find data (optional)

Contributing

1. Create a branch
2. Add/update pack content
3. PR with description of changes

Keep packs focused — better to have multiple small packs than one bloated pack.