04-naming defines force_refresh_every_n as a full re-inject forced every N prompts; the verification matrix says the rules-loader refresh is event-driven and nothing counts prompts.

04-naming defines force_refresh_every_n as a full re-inject forced every N prompts; the verification matrix says the rules-loader refresh is event-driven and nothing counts prompts.

A premise contradiction (severity minor) found by the cross-document consistency pass.

• End A: .abcd/development/brief/02-constraints/04-naming.md:208 — "Every N prompts, the prompt-router hook forces a full re-inject regardless of dedup signature match"
• End B: .abcd/development/brief/06-delivery/02-verification-matrix.md:73 — "Nothing counts prompts"

04-naming's reserved-vocabulary row says 'Every N prompts, the prompt-router hook forces a full re-inject regardless of dedup signature match', with a default of 15. The matrix's Rules-loader dedup row says the refresh is event-driven rather than turn-counted, SessionStart and PreCompact being the two events that clear the ledger, and 'Nothing counts prompts'. A prompt counter that forces a re-inject and the absence of any prompt counter cannot both describe the loader.

Evidence: .abcd/work/reviews/2026-10-26-consistency/00-summary.md, finding 107.