corpus-guide — machine-fetchable navigation for the open corpus

This repository is the structured, machine-readable map of the open Spectral Branding research corpus — Spectral Brand Theory (SBT), Organizational Schema Theory (OST), and the meaning/meaningfulness program. It is the data layer behind the human guide at spectralbranding.com/guide and the narrative map at github.com/spectralbranding.

Everything here is a generated projection of the corpus substrate — do not hand-edit. Each file is regenerated from source and screened for public safety at emit time.

File	What it is
corpus-map.json	papers ↔ terms ↔ claims navigation graph for seed → scope expansion: from a seed of 1–2 papers, widen to every paper bearing on a task via shared terms and corpus-internal citations. Navigation only — claim/method text lives in each paper's paper.yaml and the decision router.
guide-routing.json	the role map — 12 cohort accelerators (CMO, COO, CFO, researcher, AI agent, …), each with seed papers, instruments, the decisions it can act on, and an expand_via neighbourhood. An accelerator, not a gate: every role carries the other fallback.
articles-map.json	published long-form article ↔ corpus-paper edges, derived from trusted signals only (publication date ≤ today + a live canonical URL + in-body DOI references).

1 | Getting Started

Point your agent at corpus-map.json + guide-routing.json, give it your task, and let it widen from the seeds before answering — and, like the corpus instruments, tell you what it cannot resolve instead of inventing an answer. See AGENTS.md.

Raw URLs (fetch these directly):

• https://raw.githubusercontent.com/spectralbranding/corpus-guide/main/corpus-map.json
• https://raw.githubusercontent.com/spectralbranding/corpus-guide/main/guide-routing.json
• https://raw.githubusercontent.com/spectralbranding/corpus-guide/main/articles-map.json

Where this fits:

• Human guide (rendered): spectralbranding.com/guide
• Narrative org map + repositories: github.com/spectralbranding
• Decision router (semantic): consult.orgschema.com
• Measure a brand: the Brand Spectrometer — for a measurement, route to the instrument, not the papers.

2 | Project Layout

corpus-guide/
|-- README.md            <- this file
|-- AGENTS.md            <- how to point an AI agent at the map
|-- CITATION.cff         <- machine-readable citation
|-- LICENSE              <- MIT (code: reproduce.sh)
|-- LICENSE-data         <- CC BY 4.0 (JSON data artifacts)
|-- corpus-map.json      <- papers <-> terms <-> claims navigation graph
|-- guide-routing.json   <- 12-role cohort accelerator map
|-- articles-map.json    <- article <-> corpus-paper edges
|-- reproduce.sh         <- artifact validation (see section 3)
`-- output/
    |-- figures/.gitkeep
    |-- tables/.gitkeep
    `-- logs/.gitkeep    <- validation run log lands here

3 | Quick Start

The seed → scope reading recipe (the map is an accelerator, not a gate — seeds are entry points, not answer sets):

1. Start from your seed paper key(s) in corpus-map.json papers[].
2. Read each seed's terms{owns, imports, refines}.
3. For each term, look it up in terms[] and follow papers{owns, imports, refines} to every other paper that uses it — those are in scope.
4. Union citation_edges where from/to is a seed (who cites / is cited by it).
5. Ground the answer in the widened set, and state any paper you could not resolve.

To validate the shipped artifacts locally:

./reproduce.sh                  # validate all JSON artifacts, log to output/logs/
./reproduce.sh --check-only     # verify dependencies only

Note on regeneration: the three JSON files are generated projections of the private corpus substrate and are regenerated upstream at emit time (with a public-safety screen). They are not rebuilt from sources inside this repository; reproduce.sh validates their integrity (well-formed JSON, expected top-level structure) rather than regenerating them.

4 | Dependencies

• bash + git
• jq or Python >=3.9 (either one is sufficient for JSON validation)

No package installation is required; the data artifacts are plain JSON.

5 | Script Map

Script	Purpose
reproduce.sh	Validate the three JSON artifacts (well-formed, non-empty, expected top-level keys); write run log to output/logs/validate_run.log

There are no generation scripts in this repository by design — generation lives upstream with the corpus substrate (see section 3).

6 | Citation

Machine-readable form: CITATION.cff. GitHub renders a "Cite this repository" widget from that file.

7 | Licence

Dual-licence discipline:

• Code (reproduce.sh) — MIT, see LICENSE
• Data (corpus-map.json, guide-routing.json, articles-map.json, docs) — Creative Commons Attribution 4.0 International (CC BY 4.0), see LICENSE-data

---

Last updated: 2026-07-03