Dogfood-surfaced improvements: provenance everywhere, scoring contract, YMYL type, report-kit @type - #4
Merged
Merged
Conversation
… scoring contract, YMYL, report-kit @type) Ran the whole package end to end on a real site; these close the gaps that run exposed. - Provenance is now intrinsic to all 7 dimension sub-skills (geo-citability, -content, -schema, -technical, -platform-optimizer, -brand-mentions, -crawlers), not just the orchestrator — so standalone runs keep it. - geo-audit Phase 2 return contract: subagents return the orchestrator's named dimensions + a provenance tag and must NOT invent their own blended sub-composite/weights; geo-ai-visibility now returns AI Citability and Brand Authority as two separate scores. - New business type "Professional / financial services (YMYL)" in geo-audit, with its schema (FinancialService/ProfessionalService/Person) + E-E-A-T expectations + an entity-reconciliation step (legal vs trading name); reconciliation also added to the audit checklist. - audit-report-kit: generateOrgJsonLd takes a `type` (Organization/LocalBusiness/ FinancialService/ProfessionalService) so the deliverable's schema matches the audit's recommendation; report `summary` is now a field, not hardcoded boilerplate. Both default to prior behaviour — committed sample PDF + JSON-LD unchanged (verified). validate.sh passes 9/9; audit-report-kit typechecks; harborview.jsonld byte-identical. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Four improvements surfaced by actually running the package end to end on a live domain (a real GEO audit dogfood). Each one closes a gap the run exposed.
Changes
[scan]/[partial-scan]/[heuristic]/[unmeasured]discipline (and "emit—, not a number") is now in all 7 dimension sub-skills' output sections — previously only the orchestrator carried it, so standalone runs lost it.geo-auditPhase 2). Subagents must return the orchestrator's named dimensions + a provenance tag, and not invent their own blended sub-composite or weights.geo-ai-visibilitynow returns AI Citability and Brand Authority as two separate scores (the composite weights them 25% vs 20% — the run showed the sub-skill was returning one blended "58/100" that didn't map).FinancialService/ProfessionalService/Person), E-E-A-T expectations (displayed credentials, regulatory IDs), and an entity-reconciliation step (legal entity vs trading/brand name) — the audit table had no YMYL type, yet that drove the biggest findings on the test site. Reconciliation added to the checklist too.audit-report-kit:generateOrgJsonLdtakes atypeso the deliverable's schema matches what the audit recommends (a financial firm →FinancialService, not bareOrganization); the reportsummaryis a field instead of hardcoded boilerplate.Safety
audit-report-kittypechecks; the committed Harborview sample PDF +examples/harborview.jsonldare byte-identical (both new fields default to prior behaviour). Verified.scripts/validate.shpasses 9/9 (incl. the 7 new sub-skill links resolving + leak-check).🤖 Generated with Claude Code