Scope Backlog — Master Parked & Ongoing List

Last updated: May 28, 2026 · 8:25 PM EDT · Session 39 cont (MIS v2 marathon)
Purpose: Everything you've mentioned that's not currently firing as a rhythm. No blind spots. No things that fall to the wayside. Scan weekly.
Note: the LIVE/PARKED tables below are stale (last maintained Session 13, Apr 23) — see KB cleanup 38-K3 to refresh. Newest loops are in the Session 39 / Session 38 blocks below.

🟢 SESSION 2026-06-21 (Worker Bee, pre-travel) — loops captured

| 21-8 | Hardcoded "moving-target" data audit (2026-06-22) — live surfaces only | Claude | 🔴 home.html Cash tile manual array (4 debt figs) · 🟡 mildred.html KNOWN bookings (refreshed from email, needs Hospitable API) · 🟡 comms.html PEOPLE (needs contact registry). Rule: anything money/booking/person not pulled live = rot risk → wire to source. |
| 21-9 | Obligations data model — CORRECT it (Sam 6/22, locked facts) | Claude (sheet + Cash tile) | Track outstanding balance AND monthly payment as SEPARATE fields per debt. Entity fixes: US Alliance HELOC ≠ Fifth Third (amortized mortgage) — don't conflate. The 3 mortgage MONTHLY PAYMENTS = 5th Third $3,312.59 · Shellpoint $3,161.51 · Selene Finance $3,379.91 (NOT a 2nd Shellpoint; NOT balances). Include credit-card minimum payments. The home.html Cash tile manual array currently mislabels payments as balances + conflates HELOC/mortgage + misattributes Selene→Shellpoint + omits 5th Third — FIX when the sheet is built. |

🔴 SESSION 2026-06-19 (Bee + MIS, remote/phone) — open loops from the portfolio/post-mortem thread

🟢 SESSION 43 (2026-06-12, Fri) — /doctor MCP cleanup, 2 servers parked

/doctor flagged 4 MCP issues. GitHub MCP → explainer shipped (outputs/2026-06-12_12-53_research_github-copilot-mcp-explainer.html), decision pending (enable w/ PAT vs disable; recommend disable — redundant with gh). Two parked for later follow-up:

🟢 SESSION 42 (2026-06-05, Fri pre-Shabbos) — bot + comms + LevSMS loops captured

Sam's long voice-stream of asks. Bot relay bug FIXED + deployed this session (@86/@87: brain now actually sends relays to chanie/mildred/family + recognizes "talking TO a person" → relay + one-liner, no task-ify). Open BUILD loops below — none built yet (verify-first / post-Shabbos):

🟢 SESSION 40 cont 3 close-out (2026-06-02) — new loops captured

🟢 CHANIE PORTAL v-next (parked 2026-05-31) — she's now logged in to ops.hookstreetservices.com via OTP

Cloudflare Access fixed 2026-05-31: portal uses One-Time PIN (email code), app "Hook Street Ops", "Family" allow-policy = chanietreitel@gmail.com. Code emails come from noreply@notify.cloudflare.com (was in spam). She's in. Sam parked the following ("we'll work on this; in the meantime that's that") — NOT to build yet, capture only:

🟢 SESSION 39 cont (2026-05-28) — MIS v2 marathon: Stages 12 → 13c + Peek + Performance + Speak + Deal Snapshot

Shipped + committed (workspace + MIS + ops-api Worker, ~20 commits today):
- MIS v2 @37 — 210 KB Code.gs. Stage 11 institutional risk (Sharpe/Sortino/Beta/MaxDD per holding + portfolio Sharpe/Beta/HHI/VaR/CVaR + stress test + backtest). Stage 12: 15-ticker covariance MC VaR + 12-1 factor scores + cost-aware backtest + Schwab live-quote scaffold. Stage 12.5 hotfix: dashboard CSS Malformed-HTML root cause (font-family quote collision + color-scheme meta), wash-sale 30→31 days, substantially-identical SPY/IVV/VOO map. Stage 13a: Position Return card (Sam's old-sheet baseline), universe expander (added 15 held tickers, no more "2 tickers"), Override Tracking (matched-vs-overrode hit rate), Position_Costs tab + SETCOST + LOADCOSTS + IMPORTBROKER (Fidelity/Schwab CSV heuristic). Stage 13a-c: fn=peekoverlay endpoint, fn=performance endpoint, Performance card (am-I-making-money headline + top winner/loser + decision hit rate). Stage 13c: fn=expand auto-chains fn=finnhub + fn=fse.
- ops-api Worker (multiple versions): MIS overlay (replaces hard-coded 2-ticker universe check — now reads from MIS v2 fn=peekoverlay for held/verdict/cost/PnL), real after-hours / pre-market via Yahoo v8 intraday chart includePrePost (Yahoo v7 blocked), daily history via Yahoo v8 chart range=2y (Finnhub free-tier candle dead since 2024) — 3D/10D/13W/52W/MA50/MA200 all populated, volume 30d-avg ratio, 3 news headlines via Finnhub /company-news, SPY benchmark for stock-specific vs sector-beta read.
- Peek HTML (outputs/peek.html): sparkline SVG (labels OUTSIDE = no clip), comma + B/M/T cap format, type=search + enterkeyhint, clear-on-input, session badge (REGULAR/AFTER-HOURS/PRE-MARKET/CLOSED), Finnhub logo in header, volume chip (HOT/busy/normal/quiet), SPY-today + vs-SPY chips, news card (3 linked headlines), Set up Trade CTA → MIS Entry Pad, All Positions CTA → MIS dashboard, colors lightened to "hint light grey" per Sam ask.
- Universal Speak library (outputs/lib/speak.js): one drop-in, every portal page gets floating 🔊 button, tap = read aloud, tap-while-speaking = stop, long-press = voice + rate picker (en-US voices, 0.5×–2.5×), persisted to localStorage. Injected on home, peek, obligations, portal index.
- Deal Snapshot sales page (outputs/services/deal-snapshot.html): RE underwriting offer, $2,500 flat, 24h turnaround, Stripe Payment Link placeholder (Sam-side action: paste from Stripe dashboard, 2-min one-time).
- Stage 12 → 20 roadmap doc (outputs/2026-05-28_17-53_strategy_mis-v2-stage-20-roadmap.html): twin-engine doctrine (production + psychology + formatting), Option C aggressive sequence, multi-tenant SaaS destination.

Open loops headed into next session:

Strategic asks parked (not actioned this session):
- $300K → $2M trajectory (Sam stated this session). Real money needs real deployment cadence; MIS performance card will measure progress in months not days.
- Margin + compounding strategy (Sam mentioned). Adding margin to MIS v2's capital allocation needs the Plaid cash + Schwab balance feed.
- IBKR vs Fidelity vs Schwab consolidation question (Sam mentioned). Per MIS_FSE_ARCHITECTURE.md: IBKR ruled out (gateway needs always-on local machine). Stay with Fidelity (Z29720600, Z29835692) + Schwab (...898). Consolidating into one isn't worth the friction.
- DRIP impact on cost basis. Fidelity DRIP buys fractional shares at the prevailing price — those don't have a separate Position_Costs entry. Current avg cost from broker emails is weighted; Sam-managed override in Position_Costs overrides only if set.

🟢 SESSION 39 (2026-05-25) — bot brain + direct line + automation cadence

Done + committed (workspace through b748089, skills 8f0ef75): conversational bot brain (memory via Transcript tab + cross-tab actions + dedup); scripts/inbox.ps1 direct read/write line (INBOX_SECRET in gitignored config); STATE + REVIEW commands; auto-assemble + dedup trigger (dormant until setupAutoAssemble); processSamDump backfill (7 cards + groceries); command-inbox-engine skill updated to the live brain; Weekly Review cloud routine (Sun 7pm) + Monday-Eve→Tomorrow-Plan cloud routine (fires Mon 7pm, incl. tonight); Tuesday plan briefing on portal. New memory: feedback_seek_capability_not_walls. Cash history found: HookStreet-Business-OS/2026-04-29…transactions.csv (Rocket Money, 22,476 rows, 2020-2026).

Open loops:
| # | Item | Owner | Next step |
|---|------|-------|-----------|
| 39-1 | PWA app icons — only Start Here installed with a working home-screen PNG; ops portal + obligations PWA have NO separate icon | Claude (build) | add apple-touch-icon + manifest icons per PWA so each installs as its own app |
| 39-2 | Wife (Chanie) access — install the PWAs on her phone/browser; maybe a Chanie-only view | Sam/Claude | decide her view scope; share install links once icons fixed |
| 39-3 | Google Workspace / Gmail login for the PWAs + ops portal (replace Cloudflare Access) | Sam/Claude | scope auth migration (CF Access Google login vs Supabase auth) |
| 39-4 | Other-PWAs brainstorm — "easy for anyone to use" menu | Claude | propose set (grocery PWA, weekly-review view, Chanie CFO view…) |
| 39-5 | Native app / monetization — full private multi-user app (App Store), each person own login + own backend, chargeable / free-trial / onboard others ("I need to make money") | Sam (strategic) | big decision; revisit AFTER the spine + runway layer prove out — do NOT build speculatively |
| 39-6 | Runway / depletion layer — wire real cash IN (Chanie 5609 + Eden + STR) vs OUT into the brief; Rocket Money CSV is the OUT history | Claude (build) | needs Sam to confirm income source → runway line in morning + weekly brief |
| 39-7 | Migration: entire "2 Hook St." workspace → Google Drive as source-of-truth/memory; work via Claude Code sessions | Sam/Claude (direction) | plan Drive-as-truth migration; affects how cloud routines read |
| 39-8 | Obligations-sheet read into the morning brief (flagged "later increment" in buildMorningContext_) | parallel session's lane | add buildObligationsDigest_ reading v1 1L_rxCSO… |

🟢 SESSION 38 (2026-05-24) — Monetization dig + skills build

Done + committed (workspace 5a56be3, skills 2503ecf): archive deep-dive → offer defined (outputs/2026-05-24_13-05_strategy_your-offer-defined.html); v0.1 underwriting engine runs on a fake deal (tools/underwriting/hsc_underwriter_v0.py); 8 deliverables incl. 4 new/rewritten skills (underwriting-method, operations-systems-build, passover-program-ops, mis-daily-report rewrite, eden-gardens-status refresh); client services menu; coaching-chat primer; KB health audit. New memories: user_sam_professional_history, user_sam_cashflow_diagnosis, feedback_no_flattery_results_only.

Open loops:
| # | Item | Owner | Next step |
|---|------|-------|-----------|
| 38-1 | React to offer pricing (Quick Screen / Full UW / +DD / Retainer) | Sam | too high / low / right |
| 38-2 | Name 3 warm operators for outreach | Sam | from contacts (Investor/Client tags) |
| 38-3 | Contacts warm-operator shortlist | Claude | pull contacts sheet, filter, hand list |
| 38-4 | v1 underwriting engine | Claude (build) | full waterfall edge cases + snapshot PDF export + openpyxl read-back from 61 Park + intake-gating |
| 38-5 | Pick ONE vertical + run one low-risk recurring-client experiment while runway holds | Sam | RE underwriting (most proof) vs systems/ops (most energy) |
| 38-6 | Google Drive deep dig — DBI + Hook Street dropbox (Syndicator Analyzer) | Claude (optional) | deeper extraction if Sam wants |
| 38-7 | Coaching-chat primer | Sam | paste into Claude app (car); draft + outputs/2026-05-24_15-40_primer_coaching-chat-portable-context.md |

KB-health cleanups (full report: outputs/2026-05-24_15-50_audit_kb-health-check.md):
| # | Item | Fix |
|---|------|-----|
| 38-K1 | Eden #20028 "days outstanding" stated 4 ways (28 / 56 / ~79) | reconcile to ONE figure computed from invoice due date; treat BOS Operations Summary as canonical, don't hardcode |
| 38-K2 | CLAUDE.md status block stale (Session 37) | refresh to Session 38 |
| 38-K3 | CONTEXT.md badly stale (Session 36 + self-noted gaps) | full session-log refresh |
| 38-K4 | hookstreet-skills README lags (shows ~6 skills, omits new ones) | update table + diagram to 14 skills |
| 38-K5 | .remember/now.md empty (0 bytes) despite being a mandated session-open signal | investigate broken write |
| 38-K7 | No real self-learning loop — it's manual + .remember auto-capture is broken (now.md 0 bytes); Claude.ai chat can't be current (no live access + stale training). Sam: "the gap is costing, at the least." | (1) fix the .remember writer; (2) wire KB health-check into session-OPEN (not on-demand); (3) get Phase 2 bot keyed as the live-current mobile surface; (4) usage rule: Claude.ai chat = thinking only, not current facts. Memory: project_self_learning_loop_gap. |
| ~~38-K6~~ | ✅ CLOSED 2026-07-02 (VM portal-rebuild session) — scripts/build-index.py auto-generates everything.html + index.json from real dates (filename stamps + non-mass git commits); 🟢 NEW self-expires after 7 days; home's NEW strip reads index.json. Hand-maintained index rot is dead. | Run python3 scripts/build-index.py before any portal deploy (sessions do this; TODO fold into the auto-deploy when deploy-outputs.yml moves to CF Pages). |

🔴 LIVE — In Progress Right Now

📋 SHORT-TERM PARKED — This Week / Next

🏛 MEDIUM — Month-Out Parked

🌱 LONG-TERM — Parked for After Eden Closes

📑 TAX DOCUMENTATION (New Track — Apr 23)

🔁 ACTIVELY FIRING — Automated Rhythms

For reference — these are running so you know what you're NOT carrying manually.

Daily (weekdays):
- 📈 MIS Morning Brief ~9:55 AM
- 🚨 MIS 2:05 PM Trade Action
- 🚀 MIS 3:00 PM Power Hour
- 📊 Hookstreet OS Daily Snapshot ~7:40 AM

Weekly:
- 🗓 MIS Sunday Week-Ahead — Sun ~6 PM
- 📊 Hookstreet OS Weekly Review — Sun ~7:53 AM
- 📊 Hookstreet OS Operations Summary — Sun ~11:26 AM ⭐ (this is the goldmine)
- 📊 Hookstreet OS Monthly Overview — Sun
- 👩 Family Weekly Snapshot → Chanie — Sun (preview to Sam first)

SOP:
- MORNING_RHYTHM.md — daily SOP for Sam + Mildred across NY + EAT

🛠 GITHUB REPOS — cleanup queue (from docs/GITHUB_REPO_AUDIT_2026-05-19.md)

💡 FUTURE BUILD IDEAS — parked

🔬 MIS DEEP-DIVE + DATA LAYER — scoped from the 2026-05-19 voice dump

Big coherent body of work. Four clusters — each is a focused session, not a task.

Cluster A — MIS deep-dive & report
- Review the last MIS email + confirm what Codex did with the GitHub issues (all / some / none).
- Produce Claude's own MIS report (the 3 PM Power Hour equivalent) from the live sheet.
- Compare Claude's report vs the real 3 PM Power Hour — aesthetics, formatting, and the actual data/logic.
- Full audit of the 3 PM Power Hour — what's left, what Codex didn't reach.
- Independently re-derive the sheet's logic — find mistakes / wrong logic; check Sam is even looking for the right inputs.
- A "MIS from X% → 100% useful" setup plan: time+money management, momentum bang-for-buck, markets-up vs markets-down playbook, ETFs vs single stocks (when to play / not), how to use the sheet.

Cluster B — Data layer & security
- Secure the Schwab API keys / secret IDs out of the sheets (Script Properties pattern, like INBOX_SECRET).
- Schwab API: can it quote every held ticker, or are there rate-limit holes? Free alternatives?
- IBKR option — Sam has an account, willing to ~$100 for reliable data + auto buy/sold feed.
- Auto-track buys/sells — via broker API or by parsing trade-confirmation emails — sheet self-updates.
- A one-tab, view-only review snapshot for Mildred (she confirms / flags, doesn't edit).

Cluster C — Mildred's sheets
- Review all recent emails from Mildred — she sent sheets she's building; one is Contacts.
- Deep-dive the Contacts sheet; extract + reorganize. Fold her other 2 sheets into the workspace.
- Tie into Hook Street Services website.

Cluster D — Trigger reset + version retirement
- After fixing the BOS/MIS duplicate-email triggers (Delete-All / Create-Default), decide which
old sheet VERSIONS to keep vs archive — deleting triggers stops dupes, it does NOT rebuild data.

Read-before-doing (per receipts protocol): MIS folder (MIS_FSE_ARCHITECTURE, MIS_REVIEW,
MIS_TAB_AUDIT, project_mis_* memories), docs/, the live MIS sheet, last MIS email, GitHub issues.

📱 LevSMS — forward items (parked 2026-05-26, Session 39 cont.)

Context: live SMS testing 5/26 surfaced UX gaps. Hotfix shipped @34 (report-my-zip + accurate
out-of-area set confirmation). These are the NEXT-up items, in Sam's priority order:

• L1 — Set area by NAME, not just ZIP (Sam: "very important"). People don't know their ZIP;
  they know "North Woodmere" / "Five Towns" / "Lakewood." Build a curated NAME→ZIP alias map for
  the top Orthodox communities (doubles as the 50-state seed). Fuzzy match + fallback to ask.
  GATE/RISK: wrong name→location = wrong zmanim, which violates the data-integrity doctrine —
  build the map carefully + verified, not rushed. Recommended next build.
• L2 — 50-state Orthodox community schedules. Agents research shul/candle schedules per known
  kehilla. zmanim/weather/Shabbos already work for ANY US ZIP live; only SHUL-times need a loaded
  schedule. Mass-adding shul coverage collides with Gate 1 (10 NW users) + Gate 3 (2nd nbhd after
  90 stable days) — do deliberately, gated.
• L3 — WELCOME asks for ZIP/area up front. Lives in the SYSTEM_RESPONSES sheet, so a code
  default change won't take live — needs a sheet edit (or a seed-missing-keys upsert in code).
• L4 — Engagement-gated contribution. After N interactions, invite "add your shul — email
  support@levsms.com." Earned-depth, post-Gate-2.
• L5 — Premium opt-in monetization. 7-day trial → monthly; private SMS assistant (reminders,
  grocery/scheduling, restaurant recs, learns the user's style). Gated by Gate 5 (12mo + 100
  retained). Brainstorm only.
• L6 — International: Israeli Twilio number + WhatsApp channel + Hebrew for certain areas.
  Out of v1 scope (CONTEXT.md Part 19). Separate build; research cheap A2P/number approval cost.
• L-cost — Twilio billing flag. Auto-recharge ON to card 9405 (= acct 1260, the chronic-
  overdraft one), target balance $20, last top-up $10.07 on 5/26. If 1260 overdrafts, recharge
  fails → service interruption. Consider moving to a stable card. (Personal inbox has no Twilio;
  business only.)
• L7 — Krias-Shma "repeat" — SHIPPED v1 (@42). Out-of-area Maariv line + NW Maariv response now
  note "if you daven Maariv before tzeis, repeat Krias Shma after tzeis." Morning Shema/Tefila
  deadlines + proximity alert shipped @38. REMAINING (future): the per-user OPT-IN reminder (store
  preference, proactively remind) — needs opt-in infra; informational v1 covers the common case.
• L8 — Multi-timezone — SHIPPED (@42). fmtTime_/nowMins_/formatZmanim/approxDavening now render
  in the ZIP's real zone via Hebcal location.tzid (was Eastern-only via getHours → CA was 3h off).
  Works for ANY US ZIP now. REMAINING: Israel (different postal system + Hebrew) = L6.
• L9 — Sponsorship "brought to you by" footer. ZIP-coded, disclosed (per Trust Philosophy:
  every sponsored placement labeled). Sam's "later thought" = MONETIZATION, deferred. PILOT STAND-IN
  SHIPPED (@41): a soft pilot CTA (feedback/help/intros/share via support@levsms.com) occasionally
  closes content replies, <=1/user/day, rotating, GSM-7-safe. Swap to real sponsorship at Gate 5.
• L10 — Incorporation trigger + business bank account. Form Malca Yenta Networks LLC + EIN +
  a SEPARATE business account BEFORE the first revenue dollar (sponsorship/premium) — NOT needed
  for the free pilot. Codex's DE filing packet is ready to execute. Money in → separate account,
  never commingled (commingling pierces the LLC liability veil).
• SHIPPED: L1 (set area by name, @36); zmanim-aware out-of-area davening (@38); GSM-7-safe
  footer (@38); "North Woodmere"->"NW" + framework msgs code-managed (@39); time-aware NEXT prayer
  (morning shows next Shacharis w/ shul, not always Mincha, @40); pilot CTA (@41);
  L8 timezone-correct anywhere incl. California (@42); L7 repeat-Krias-Shma note (@42).

🧭 How to use this file

• Weekly scan: Sunday morning after Ops Summary arrives, scan this file. Move items: live→short-term, short-term→done, unblocked M# items up the list.
• When you remember something: drop it here first as a one-liner. Categorize later.
• When you finish something: check it off and commit.
• When something stalls: capture WHY it stalled so future-you knows the unblock trigger.

Each row is a one-line decision or one-line next-step. If something takes more than one line, it belongs in its own .md file, linked from here.

Updates since creation:
- Apr 23, 2026 (Session 13) — created during Phase 1 Eden close-out. 43 items at creation.
- May 19, 2026 — added GitHub repos cleanup queue (G1–G4) from GITHUB_REPO_AUDIT_2026-05-19.md.

MIS operator open loops (2026-06-19, session f527be02)

• Wire misV2ImportTrades_ to auto-run — MIS holdings go stale because trades never import; this is the root of the 'you're not seeing my trades' frustration. Pending Sam go. (MIS lane)
• META concentration ~50% of equity (margined) — trim plan offered, pending Sam choice. (MIS lane)
• Lock reconciled live book into HOLDINGS_CLEAN (META ~18.8sh, IBIT 10, MU flat, ABBV exited, SpaceX Joint exited/son holds). (MIS lane)
• MRVL discrepancy — Sam says he sold Marvell; no broker email exists. Clarify. (MIS lane)