Base 16 v3 — Opus 4-6 Per-Seed Breakdown

16/16 seeds. Gen + Judge: claude-opus-4-6. ✅ = pass, ❌ = fail, — = N/A (root seed, no parent)

Judges: SV=SeedValue, PS=PromptSpecificity, CI=ContractIndependence, CC=ContractConcreteness, Nav=Navigability, TD=TokenDensity, SC=SlugCompression, Dst=Distinctiveness, Grm=Grammar, RC=ReferentialClarity, PC=ParentChildCoherence, VA=VerificationAccuracy

Seed	→ Generated slug	SV	PS	CI	CC	Nav	TD	SC	Dst	Grm	RC	PC	VA	Tot
📧 Email
gmail.api.ready	gmail.auth.preflight	✅	✅	✅	✅	✅	❌	✅	❌	✅	✅	—	✅	9/11
gmail.inbox.health	inbox.health.diagnostic	✅	✅	❌	✅	✅	❌	❌	❌	✅	✅	❌	✅	7/12
gmail.search.query	gmail.search.query	✅	✅	❌	✅	✅	❌	❌	✅	✅	❌	❌	✅	7/12
gmail.thread.summarize	thread.synthesize.brief	✅	✅	❌	✅	✅	❌	✅	❌	✅	✅	❌	✅	8/12
gmail.writing.analysis	voice.profile.inbox	✅	✅	❌	✅	✅	❌	✅	❌	✅	✅	❌	✅	8/12
gmail.draft.compose	ghost.writer.draft	✅	✅	❌	✅	❌	❌	❌	❌	✅	❌	❌	✅	5/12
📁 Workspace
local.files.ready	filesystem.readwrite.check	❌	✅	❌	✅	✅	❌	✅	❌	✅	✅	—	✅	7/11
local.project.scan	project.fingerprint.detect	✅	✅	❌	✅	✅	❌	✅	❌	❌	✅	❌	✅	7/12
document.file.selected	file.intake.verify	✅	✅	❌	✅	✅	❌	✅	❌	❌	✅	—	✅	7/11
document.structure.map	document.structure.map	✅	✅	❌	✅	✅	❌	✅	✅	❌	❌	❌	✅	7/12
👤 Person
context.person.select	person.focus.lock	✅	✅	❌	✅	✅	❌	✅	❌	✅	✅	—	✅	8/11
person.web.profile	person.profile.briefing	✅	✅	❌	✅	✅	❌	✅	❌	❌	✅	❌	✅	7/12
🖼️ Image
image.edit.model	image.edit.ready	✅	✅	❌	✅	✅	❌	✅	❌	✅	✅	—	✅	8/11
image.user.portrait	portrait.intake.validate	✅	✅	❌	✅	✅	✅	✅	❌	❌	✅	❌	✅	8/12
headshot.professional.edit	portrait.studio.headshot	✅	✅	❌	✅	❌	❌	❌	❌	✅	✅	❌	✅	6/12
image.style.transfer	portrait.style.identity	✅	✅	❌	✅	✅	❌	❌	❌	✅	❌	❌	✅	6/12

Aggregate pass rates:
🟢 PromptSpecificity 100% · Concreteness 100% · Verification 100% · SeedValue 94% · Navigability 88% · RefClarity 75%
🟡 SlugCompression 69% · Grammar 69%
🔴 Distinctiveness 13% · ContractIndependence 6% · TokenDensity 6% · ParentChildCoherence 0%

Systemic: CI, TD, PC fail on nearly every seed — these are conjugation model behaviors, not prompt quality issues.