Licence hygiene — the corpus gate#

Every source.yaml MUST declare a licence_bucket. Full field spec: _meta/licence-tagging-schema.md.

• green — ingest freely. OGL primary law & official gov/regulator guidance; public domain.
• amber — ingest ONLY with the specific re-use licence recorded (e.g. case law under the

National Archives computational/AI re-use licence; model docs whose terms permit corpus use).

• red — reference only, NEVER stored in this repo (Practical Law, LexisNexis, Westlaw, textbooks,

firm know-how). A human may cite them; the corpus may not contain their text.

Default when unsure: red. CI should reject any source.yaml that is red, untagged, or whose licence_bucket=amber has an empty licence_conditions.