Historic EIOPA Yield Curve Skill

A self-contained Claude Skill for answering questions about historic EIOPA risk-free rate (RFR) yield curves — spot rates, full term structures, tenors over time, country comparisons, implied forwards, Solvency II stressed curves, and curve parameters (UFR, LLP, alpha, CRA, VA).

No network access, database, or third-party packages required. Every monthly EIOPA release is bundled directly into this repo as a compact binary pack (~370 KB), so lookups run offline in well under a second.

historic-eiopa-yield-curve/
├── SKILL.md                    # Skill instructions Claude reads to know how to use this
├── references/
│   └── dataset.md              # Currency/country label table, pack format, shock formulas
└── scripts/
    ├── eiopa_curves.py         # The lookup tool — reads the pack and nothing else
    ├── data/
    │   └── curves.pack         # Bundled snapshot: every EIOPA release, all curves, terms 1–150y
    └── tests/
        └── test_eiopa_curves.py

• 140 monthly releases, 2014-12-31 through 2026-07-31 (as of this snapshot)
• 53 distinct curves, terms 1–150 years, whole-year nodes
• Both no_VA and with_VA curve types
• 54 countries/currencies, including all EEA members plus a set of non-EEA curves (dropped from EIOPA's publication starting 2025-01-31 — see coverage --country XX for per-country availability)

This is a snapshot, not a live feed. It cannot know about EIOPA releases published after the snapshot date. Run coverage or facts (see below) for the exact cutoff before relying on "latest."

1. Clone this repo, or download it as a zip.
2. Copy the historic-eiopa-yield-curve/ folder into the skills directory your Claude client reads from. (This path differs by product — Claude Code, Claude Desktop/Cowork, and claude.ai each have their own convention; check your client's docs for where custom skills live.)
3. That's it — no build step, no dependencies beyond Python 3. Claude will pick up SKILL.md and know when to use it.

You can also run the tool directly from the command line without Claude at all:

python scripts/eiopa_curves.py spot --country Germany --term 10 --date 2024-12-31

Add --va for the with_VA curve (default is no_VA), --df for discount factors, and --json for machine-readable decimal output. Full details, caveats, and known traps (one curve per currency, not per sovereign; shifting country/currency labels over time; non-constant coverage) are documented in SKILL.md.

EIOPA publishes a new release monthly. To refresh the bundled pack against a fresh checkout of EIOPA's published data:

python scripts/eiopa_curves.py pack --data-dir <path-to-EIOPA_all_curves-checkout>/data

Run the test suite afterward — python scripts/tests/test_eiopa_curves.py — which includes a snapshot-freshness check.

python scripts/tests/test_eiopa_curves.py

Pure stdlib unittest, no pytest, no network. Pins known rates and verifies every structural claim in SKILL.md against the bundled data.

Underlying rates are published by EIOPA (European Insurance and Occupational Pensions Authority) under its Solvency II risk-free rate methodology. Check EIOPA's terms for redistribution of the underlying data before republishing this repo publicly.