# The Treasury Corner — agent data layer

> Machine-readable claim ledgers for Bitcoin treasury companies, published by Mundane Tools. Dated management claims (filings, press releases, earnings calls, executive social), classified under a frozen taxonomy and scored HIT / PARTIAL / MISS / OPEN against later disclosures. Static JSON. No API key. No tracking. Not investment advice.

## Manifest
- [Agent manifest](https://mundanetools.com/agent.json): every data resource with sha256, byte size, and record count. License CC-BY-4.0. Cite claims by immutable ID (e.g. KULR-009).

## Data
- [All claims](https://mundanetools.com/data/claims.json): every scored management claim, one record per claim
- [Claims — Strategy (MSTR)](https://mundanetools.com/data/claims/mstr.json)
- [Claims — Strive (ASST)](https://mundanetools.com/data/claims/asst.json)
- [Claims — Nakamoto (NAKA)](https://mundanetools.com/data/claims/naka.json)
- [Claims — KULR](https://mundanetools.com/data/claims/kulr.json)
- [Company score vectors](https://mundanetools.com/data/companies.json): execution / strategic / high-materiality trust decompositions — never a single composite
- [Taxonomy v1.1](https://mundanetools.com/data/taxonomy.json): the frozen classification and scoring rules
- [Predictions ledger](https://mundanetools.com/data/predictions.json): this project's own dated, scored forecasts
- [Article index](https://mundanetools.com/data/articles.json)
- [Changelog](https://mundanetools.com/data/changelog.json)

## Reading rules
- Scores are the decomposition (execution vs strategic vs high-materiality, with denominators), never a bare hit rate.
- CAUTIOUS and PROMOTIONAL claims are counted but never resolved as hit/miss. Refinancing never scores as deleveraging.
- Each record carries a source_class (primary / mirror_verbatim / secondary) — filter on it if provenance matters.

## Human site
- [The Treasury Corner](https://mundanetools.com/): essays, methodology, and the predictions ledger in HTML