Imported from docxology/template (
projects/templates/template_methods_paper/data/AGENTS.md). Install upstream withnpx skills add docxology/template --skill data. Copyright stays with the author.
Data Directory — Agent Guide
Versioned project inputs only. Pipeline outputs must not be committed here.
This exemplar has no external dataset: its "input" is the controlled
vocabulary declared in code (src/template_methods_paper/methods_dsl/vocabulary.py,
src/template_methods_paper/methods_dsl/units.py), not a CSV fixture. data/ therefore holds only
the claim ledger, matching template_code_project's convention for exemplars
without a dataset.
claim_ledger.yaml
Evidence-registry for manuscript claims that are intentionally sourced from
code, the dataset, or generated reports rather than {{VARIABLE}} injection.
Schema (preserve when adding rows)
| Field | Purpose |
|---|---|
claim_id |
Stable identifier |
kind |
Claim category |
value |
Declared numeric or textual value |
source |
Provenance (module, manuscript section, artifact) |
source_tier |
Trust tier for validation |
freshness |
Staleness policy |
artifact_path |
Optional path to backing file |
Edit protocol
- Edit
claim_ledger.yamlonly when manuscript claims, figure defaults, or source-backed numeric facts change. - Re-run evidence validation / pipeline stages that consume the ledger.
- Do not store generated CSV/JSON/PNG under
data/— those go tooutput/.
Quick orientation: README.md.