@@ -2,36 +2,47 @@
 
 This folder tracks remaining DebateLab product work. The source tree is the authority: a plan is active only when it describes work that is not already present on the current branch.
 
-## NOW
+## Current branch status
 
-1. **Sparring v1** — turn DebateLab from a preparation toolbox into a repeatable training loop: motion → side → opponent exchange → qualitative scorecard → targeted next drill.
-2. **Verification loop** — keep PR CI deterministic, add browser coverage for the critical product journeys, and prepare authenticated staging E2E without applying or deploying any Supabase changes from this branch.
-3. **Post-remediation security validation** — verify the August security findings against current source and add regression tests around the remediated trust boundaries.
+`feature/sparring-verification-retention-loop` implements the planned product/verification source work:
 
-## NEXT
+- Sparring v1: four-speech practice loop, transcript-grounded qualitative scorecard, retry recovery, local history, exports, analytics, and handoffs into Ask/Attack/Brainstorm/Factoids.
+- Verification loop: one `npm run verify` gate covering lint, typecheck, coverage, build, and deterministic Playwright browser journeys with retained failure evidence.
+- Authenticated staging E2E: prepared as a separate opt-in stateful suite, deliberately excluded from PR CI and not executed while remote Supabase mutation is prohibited.
+- Post-remediation security validation: original August findings mapped to current source, regression tests strengthened, and branch-only backend hardening recorded under `security/reviews/`.
+- Search Quality: first-class Exa domain/category filters, raw request tests, handler-boundary research coverage, and Factoids consulted-source provenance/UI are implemented in source.
 
-- [Ask and sitewide search quality](./search-quality.md): finish the remaining Exa domain-filter verification, handler-level integration coverage, and Factoids consulted-source UI.
-- Sparring follow-ups driven by product data: format presets, richer drills, voice/live-round experiments, and longitudinal progress only after the core loop proves useful.
-- Conversion and retention work based on measured activation and repeat usage.
+No branch migration or Edge Function has been applied/deployed to Supabase.
+
+## NEXT — data and authorized deployment, not another feature pile
+
+- Run the full verification gate on an available runner and fix any executed failures.
+- When explicitly authorized, review/apply the pending Supabase migrations and deploy changed functions to a controlled staging environment, then execute the separate stateful staging suite.
+- Run a fresh official security scan against the deployment candidate after those backend changes are actually active.
+- Measure Sparring activation, completion, rematch/revisit behavior, feedback-action usage, and conversion before expanding the training product.
+- Choose follow-up Sparring work from evidence: format presets, richer drills, voice/live-round experiments, and longitudinal progress only if the core loop proves useful.
+- Conversion/retention work should follow observed funnel behavior rather than assumed feature demand.
 
 ## LATER
 
 - [Admin announcements](./admin-announcements.md)
 - [Animation polish](./animation-polish.md)
 - [Rate-limit UI cleanup](./rate-limit-ui-cleanup.md)
-- Further model/prompt experiments that are supported by evaluation data.
+- Further model/prompt experiments supported by evaluation data.
 
 ## Historical / substantially implemented
 
 These files are kept for implementation history. Do not treat them as current instructions without checking the source tree first.
 
+- [Ask and sitewide search quality](./search-quality.md) — implementation record; formerly deferred domain filtering, handler coverage, and Factoids consulted-source UI are now present on this branch.
 - [Streaming AI](./streaming-ai.md) — Ask streaming endpoint/client/tests are already present.
-- [Source auto-import](./source-auto-import.md) — import endpoint, validation, schema, Library integration, and tests are already present.
+- [Source auto-import](./source-auto-import.md) — import endpoint, validation, schema, Library integration, and tests are present; this branch further hardens its network egress in source.
 - [Control honesty and deterministic toggles](./control-honesty.md) — core enforcement is implemented and tested.
+- [Debate sparring and streaming](./sparring-and-streaming.md) — original design record; Sparring v1 is implemented with a deliberately narrower four-speech loop.
 - [Model cost accounting](./model-cost-accounting.md) — accounting helpers, audit metadata, and tests are present; remaining work is operational validation.
 - [System prompt improvement](./system-prompt-improvement.md) — layered shared prompt architecture is present.
-- [Quality & Truth release checklist](./release-checklist.md) — use as a release-history checklist, not a live roadmap.
-- [Supabase staging](./supabase-staging.md) and [Agent verification loop](./agent-verification-loop.md) — still useful design references, but any migration/function/secret deployment is explicitly out of scope for this feature branch.
+- [Quality & Truth release checklist](./release-checklist.md) — use as release history/checklist, not a live roadmap.
+- [Supabase staging](./supabase-staging.md) and [Agent verification loop](./agent-verification-loop.md) — useful design references, but migration/function/secret deployment is explicitly outside the current branch execution boundary.
 
 ## Done