Marketplace — Feedback and Reputation

[Polyterative]

Marketplace — Feedback and Reputation

Status

Safe local helper checkpoint in progress. Priority: MEDIUM. Product area: marketplace / trust / public profiles.

Current checkpoint has pure TypeScript helper and co-located specs for future post-close feedback prompts. User selected
Feedback/Reputation as the next Marketplace slice after address book + listings and approved additive schema/RLS/backend/moderation
support. Deploy, release, push, and production-branch work remain gated on explicit approval.

User intent

The marketplace needs feedback management and a reputation system that uses transaction outcomes plus trustworthy account
activity signals such as account age and constructive contribution history.

Product / roadmap fit

This extends the existing trust-tier philosophy without turning Patcher into a status game. Reputation should help buyers
judge risk, not create leaderboards or creator loops.

Current system analysis

• Product principles allow trust tiers for UGC and reject leaderboards, streaks, and profile-ranking mechanics.
• Public profiles can display contributor stats and will be the natural identity surface for seller trust.
• Transaction lifecycle is needed before feedback is meaningful.

Future strategy

Use bounded trust signals instead of a single public score. Display context where users need it: listing cards, listing
detail, seller profile, and transaction prompts. Do not build a public reputation dashboard.

Safe local helper checkpoint

• Define MarketplaceFeedbackDraft with transactionId, giverProfileId, receiverProfileId, sentiment, and optional body.
• Add local validation/normalization for draft-shaped unknown values without throwing.
• Require nonblank transaction/giver/receiver IDs.
• Restrict sentiment to positive, neutral, or negative.
• Trim optional body text and cap it at 500 characters.
• Require nonblank body context for neutral and negative feedback; allow positive feedback without body.
• Return only { valid: true, feedback } or { valid: false, errors }.
• Whitelist normalized feedback output and exclude unknown/private draft fields.
• Record approval for schema/RLS/backend/moderation work after address book + listings foundation.

Goals

• Add feedback only for completed transactions.
• Use positive / neutral / negative sentiment, not 1-5 star ratings.
• Aggregate trust signals from completed sales, purchases, disputes, account age, and approved contributions.
• Prevent easy retaliation and sockpuppet inflation.
• Show seller trust inline on marketplace surfaces.

Non-goals

• No public leaderboard.
• No follow/friend graph.
• No open-ended profile wall comments.
• No feedback before transaction volume exists.
• No opaque numeric "seller score" in MVP.

Assumptions

• Feedback should launch after enough transactions exist to make it useful.
• Buyer and seller should both be able to leave feedback.
• Feedback text needs moderation hooks.

Dependencies and sequencing

Depends on closed transactions from marketplace-transaction-lifecycle.md. It can be designed in parallel but should not
ship with marketplace MVP unless moderation and anti-abuse rules are also ready.

MVP layer

• Propose and approve transaction_feedback schema / RLS before migration work.
• Add post-close feedback prompt in buyer/seller cockpit rows.
• Add sentiment feedback with optional positive note and required neutral/negative moderation context.
• Add feedback visibility rule: show after both submit or after feedback window expires.
• Add seller trust summary on listing detail, visible to anonymous visitors as public-safe aggregate context.

Structural layer

• Add profile_trust_signals view or query aggregation.
• Add contribution signals: account age, approved module submissions, approved price reports, completed sales/purchases.
• Add anti-abuse caps for repeated same-pair feedback.
• Add dispute-aware reputation display.
• Add moderation queue integration for feedback text.

Polish layer

• Add public profile feedback tab with bounded transaction feedback.
• Add "new seller" neutral trust band that calibrates risk without shaming.
• Add explanatory methodology copy for trust chips.

Proposed data model

transaction_feedback

• id
• transaction_id
• giver_profileid
• receiver_profileid
• sentiment
• body
• created_at

profile_trust_signals should be a view/query, not a manually maintained reputation table, at least for the first pass.

No schema/RLS/policy/migration/data change is approved by the current helper checkpoint.

Proposed RLS (requires user approval)

• Feedback insert only by transaction participant after transaction is closed.
• Feedback read follows visibility window / two-sided release rule.
• Feedback body moderation may require admin-only read/update paths.
• Aggregate trust signals public-safe only; no private transaction detail leakage.

Anti-abuse rules

• Feedback window: 30 days after close.
• Two-sided release: hidden until both parties submit or the window expires.
• Same giver/receiver pair capped over a rolling period to limit self-boosting.
• New accounts show a neutral "new seller" band instead of a negative mark.
• Disputes are represented as bounded flags, not a punitive score.
• Reputation remains attached to the stable profile, not the current username; username changes after transactions should preserve continuity through redirects/aliases.

File / surface map

• Local helper checkpoint: src/app/features/marketplace/marketplace-feedback.utils.ts
• Local helper specs: src/app/features/marketplace/marketplace-feedback.utils.spec.ts
• Buyer/seller cockpits from marketplace-browse-detail-and-cockpits.md
• Public profile route
• Listing card/detail trust chips
• src/app/features/backend/DatabaseStrings.ts
• src/app/features/backend/supabase-add.ts
• src/app/features/backend/supabase-get.ts
• src/backend/database.types.ts

Acceptance criteria

• Feedback can only be created for closed transactions.
• Feedback cannot be used for generic profile comments.
• Listing surfaces show concise trust context without a public score leaderboard.
• Aggregate trust signals do not expose private transaction details.

Validation strategy

• Unit tests for feedback eligibility, visibility windows, draft validation, body normalization, no-throw malformed inputs, and whitelisted output.
• Supabase service tests for feedback creation/query paths.
• Component tests for listing trust chips and feedback prompts.
• Abuse-case tests for same-pair caps and non-participant access where possible.

Latest safe helper checkpoint validation (2026-07-07T13:35+02:00):

• pnpm test-headless --include="**/marketplace-feedback.utils.spec.ts" passed.
• node scripts/checks/check-docs.cjs passed.
• git diff --check passed.
• pnpm lint passed.

Risks and open questions

• Decide feedback body length and moderation policy.
• Feedback/trust summaries are visible to anonymous visitors as public-safe aggregate trust context.
• Allow username changes after completed transactions, but keep reputation attached to the same profile and preserve public continuity with redirects/aliases.
• Decide which contribution stats are public-safe in v1.

Coordinator-loop handoff

Do not implement before transaction lifecycle has closed-sale data. Involve advisor/reviewer personas for anti-abuse and
privacy review before schema approval.

Current helper checkpoint may be tested and refined locally, but must not be wired into UI or backend persistence before
schema/RLS/moderation approval.

Decision log

• 2026-06-18T11:26+02:00 — Reputation planned as bounded trust context, not stars, scores, or social-status loops.
• 2026-07-06T18:02+02:00 — Added safe local feedback eligibility, visibility, and bounded trust-band helpers with specs; no schema/RLS/remote apply/UI/deploy was done.
• 2026-07-07T13:33+02:00 — Added safe local feedback draft validation/normalization checkpoint: whitelisted output, nonblank IDs, sentiment whitelist, body trimming/capping, and no-throw malformed input handling only.
• 2026-07-07T13:33+02:00 — Neutral and negative feedback require nonblank body context as the safe product default so future moderation review has explanation; positive feedback may omit body.
• 2026-07-07T13:33+02:00 — Schema, RLS, backend methods, moderation persistence/workflow, UI prompts, routes, deploy, release, push, and production-branch work remain explicitly gated.
• 2026-07-07T14:07+02:00 — User selected Feedback/Reputation as the next Marketplace backend/schema target after address book + listings and approved additive schema/RLS/backend/moderation support. First UI placement should be user profile/seller reputation areas plus completed transaction surfaces.
• 2026-07-08T14:30+02:00 — User chose public-safe Marketplace feedback/trust summaries visible to anonymous visitors, not logged-in-only.
• 2026-07-08T14:30+02:00 — User chose profile-attached Marketplace reputation continuity: username changes remain allowed after transactions, but public profile redirects/aliases should preserve trust continuity.
  Source: internaldocs/workflow/plans/marketplace-feedback-and-reputation.md
  Migration status: migrated from local documentation