S2: dest-class lists (P59 143/180)

[sdntsng]

Summary

S2 enumerate for dest-being skill sets, transfer-cue item objects, and possessed-class dest names. Same-store skip-ingest 180 on tenant diag-mh-135 is 143/180 (0.794) vs P58 141/180. Unique gains: Audrey dog items (conv-44-q51) and James pet tricks (conv-47-q40). Unique leftover losses: none.

This is not leftover covering. It is not 90% (162/180 on this sample; public LoCoMo is n=1540 at 11.4%). It is not a Mem0 same-pin. Do not replace README 11.4% / 70% 1×30.

Not-pins on this branch: p59 142/180 (Jolene snakes leftover loss) and p59b 142/180 (Maria dogs crowded with generic named David).

Product SHA: 2111b3b. Pin-docs SHA: 7582832 (FF target for dev).

Changes

• internal/memory/hop_executor.go / internal/memory/recall.go: dest-being skills recover dest capabilities (swim/catch/balance/skateboard) without requiring trick; dest-adjacent identity dumps stay out. Transfer-cue item objects (buy/acquired/purchased; got/made only with for/gift) recover collars/tags/toys/beds; coded keys and slogan-filtered transfer objects stay in when they have a transfer cue. Name lists recover dests from a possessed-class token plus named/called, not generic named X friends.
• internal/memory/recall_test.go: fixtures for transferred item sets, dest-only tricks, possessed-class names (Coco+Shadow not David; Jolene snakes not identity dumps).
• Pin docs: docs/benchmarks/README.md P59 row; docs/research/competitive/cycle-closeout.md P59 section (Landed → Own pins → Competitor compare → Why → Next); artifact locomo-s0-diag-mh-135-p59-20260826.md + summary JSON + failure ledger jsonl.

Test plan

• go test ./internal/memory/ (unit fixtures for transfer / dest skills / possessed-class names)
• Skip-ingest 180 locomo-s0-diag-mh-135-p59c-product-recall-s1-5336d8 → 143/180, MH 23/33, OD 4/11, SH 85/98, temporal 31/38; unique losses none vs P58
• Held: P54 counts, Shinjuku, leftover covering (graceful, snow-confused, soda/candy, yoga+running), Jolene snakes, Maria dogs
• gofmt -l . / go vet ./... on merge
• Do not merge leftover-covering PRs S1–S5 compiler/answer/alias/KU (split from mixed S0 PR) #133, S0–S5 SOTA execution increments (stratified harness + compiler/answer/hop/LME) #131, fix(recall): instrument-purpose leftover covering over help flood #143, P29: demote single-token attended compiler echoes (ranking only) #145

Linear / tracking

• ENG-176 (MH / S2 enumerate)
• Parent ENG-168

Risk / follow-ups

• Remaining 37: RETRIEVAL 13, PROOF 14, READER 5, WRITE 4, HARNESS 1. Next generic product: food/meal suggestion lists (conv-49-q18, conv-49-q37) without soda/candy hop dumps; then S2b OD; then S1 WRITE with re-ingest; then S5 industry (62/180).
• 143/180 is not 90% and not public LoCoMo. Fair Mem0 180 waits on quota 2026-09-01.
• Do not FF main. Do not write SOTA / beats-Mem0.