Reference Corpus

Status: Working draft

The Reference Corpus is the primary structured evidence base for Framework Engineering research.

It stores characterized knowledge artifacts using FEMS-1.

The corpus is artifact-first.

Each artifact has its own folder.

CSV files are generated indexes and summaries, not the source of truth.

Canonical artifact data lives in:

research/corpus/artifacts/<artifact_id>/

Each artifact folder may contain:

• README.md
• kacs-v1.yaml
• evidence.md
• notes.md
• characterization-history.md

Schema note: secondary_artifact_types has been replaced by capabilities in the FE-008 identity-capability update. Historical references to the older field name should be treated as legacy vocabulary.

Blind-review note: Validation corpora now record identity, capabilities, framework_grammar, and primitive_composition separately so agreement can be measured independently for each dimension.

Measurement note: FEMS-1 is the active instrument for FE-008 and future related corpus work. Schema changes should be versioned rather than silently reinterpreted.

Rules:

• Do not overwrite prior KACS versions.
• Create kacs-v2.yaml when characterization materially changes.
• Record why changes occurred in characterization-history.md.
• Maintain evidence profiles for identity, capabilities, composition, and lifecycle.
• Every artifact must identify the measurement standard used.
• Every artifact must include review metadata.