Net Zero Industrial Policy Lab – Claude Code Context

Shared context for the NZIPL team. This file loads automatically in any Claude Code session opened from this repo.

About the Lab

The Net Zero Industrial Policy Lab at Johns Hopkins University identifies and sequences clean industrial opportunities for countries using economic complexity methods. The Lab works with governments, multilateral institutions, and development finance partners to translate capabilities analysis into actionable industrial policy, with active programs focused on Brazil, India, and Mexico.

Outputs span interactive analytical products (play cards, infrastructure maps, cross-country comparators) and traditional research formats (policy briefs, working papers, methodological notes). This kit supports the Lab’s tooling and shared conventions; the research itself lives in published work and project-specific repositories.

Website: netzeropolicylab.com Platform: cice.netzeropolicylab.com (Clean Industrial Capabilities Explorer)

Research agenda

Current threads the Lab is working across:

• Clean industrial policy sequencing — which plays to develop first, in what order, with what preconditions
• Subnational industrial policy — state industrial policies in Mexico
• Subnational industrial capabilities — state- and municipality-level RCA, relatedness density, capability clusters
• Green investment — How can countries sustain investment at home to create jobs and boost domestic growth
• Finance architecture for industrial policy — instruments, actors, bottlenecks, and the coordination problem
• Cross-country comparison methods — standardizing indicators so Brazil, India, and Mexico are comparable

Team

This kit is a shared resource. As Lab members contribute, add yourself and your focus below.

How this kit fits

The kit is tooling, not a source of truth for Lab research positions. It accumulates conventions (data source auth patterns, naming norms, design tokens), reusable pipelines, and hard-won gotchas so the next person doesn’t rediscover them. Research outputs — papers, briefs, presentations, datasets — stay in their own repos and publication channels. When in doubt: can this knowledge save a colleague an hour next Tuesday? If yes, it belongs here.

Deliverable patterns

The interactive analytical products the kit currently supports. Traditional research formats (papers, briefs, conference talks, methodological notes) sit alongside these; the kit focuses on what benefits from shared tooling.

Sprint Structure

Each country follows a 5-month production + 1-month QA cycle:

Kit coverage by country

Reflects what’s in this kit — pipelines, data caches, deliverables that have been systematized for reuse. Research programs may be further along than the tooling suggests; add to this table as kit coverage expands.

Data Sources

OEC API Reference

• Base: https://api-v2.oec.world/tesseract/data.jsonrecords
• Auth: Three headers required:
  • Authorization: Bearer <key>
  • X-API-KEY: <key>
  • token=<key> as query param
• Critical: Also requires a browser-like User-Agent header (Cloudflare protection)
• Response: Records are inside a data key: use result.get("data", result) or result["data"]

DataMexico API Reference

• Base: https://www.economia.gob.mx/datamexico/api/ (legacy api.datamexico.org is deprecated)
• Endpoints: /data (cubes), /stats/rca (specialization, also the only path with full DENUE sector coverage)
• No auth required, but set a browser User-Agent
• Critical: The /data endpoint silently drops manufacturing (31-33), wholesale (43), info (51), professional services (54) from DENUE. Use /stats/rca with multi-month parameter (e.g., Month=20250522,20241126,20240523) to get all sectors.
• Full reference: install the datamexico-api skill (this folder) for the curated 14-cube catalog, 12 gotchas, and 8 stdlib templates. The skill auto-triggers when Claude needs to write or extend a Mexico pipeline.

Play Card Structure (8 Sections)

Pipeline Conventions

• Language: Python 3, stdlib only (json, argparse, urllib). No pip dependencies.
• Output: JSON files. All data embedded in HTML deliverables as const DATA = {...}.
• Caching: API responses cached locally (*_cache.json). Use --skip-api flags to reuse caches.
• Self-contained: HTML deliverables make no runtime API calls. Fully offline-capable.

Design System

Use the /nzipl-design skill for all visual formatting. Key tokens:

• Dark theme: body #1A1B1E, cards #25262B
• Brand green: #56a360 (accent), #7fc77f (links)
• Font: Archivo (headings 600-700, body 400)

• Attribution: “Net Zero Industrial Policy Lab

  Johns Hopkins University

  netzeropolicylab.com”

Skills

Commands

Tasks

References

• glossary.md – Terms, acronyms, internal vocabulary
• gotchas.md – Known data/API issues that cause silent failures
• gotchas-frontend.md – D3, headless browser, and HTML rendering issues
• discoveries.md – Team-contributed findings (append here when you learn something)
• prompts.md – Starter prompts for new team members
• pre-read.md – 5-min pre-session skim note (no install required)
• CONTRIBUTING.md – How to add to this kit