# Weekend picks — autonomous Saturday queue

Date: 2026-05-16

**Hard filters:**
- bucket == EXECUTION_READY
- owner == claude
- time_horizon in [immediate, near_term]
- supervised_execution_recommended == false

**14 tasks pass the filters.** (16 EXECUTION_READY minus 2 supervised: catalog-refactor-shared-json and add-user-role-approval.) These are ranked by tier with reversibility as tiebreaker.

> *Updated 2026-05-16 consistency pass: count was previously stated as "13" — incorrect arithmetic. The corrected EXECUTION_READY count rose from 14 to 16 after `ux-cross-nav` and `mtf-vp-tradingview-handler` were re-bucketed from a PROBABLY_SHIPPED regex misfire. Both appear as new picks below.*

---

## Pick 1 — Cron the freshness updater, then ship the banner

- **Task id:** p0-freshness-dashboard
- **Property:** foundation (INFRA / CROSS-CUTTING)
- **Time horizon:** immediate
- **Estimated effort:** S
- **Tier:** 1 compounding
- **Reversibility:** HIGH
- **Leverage justification:** Compounding across 7 active properties (foundation hosts the dashboard; scorecards/mgp/accountability_hud/param_surface/strategy_library/playbooks/live_classifier consume it). Three silent regressions in the last 9 days each cost a full trading day to detect — this closes the loop. Update_freshness.py is already written and seeded (per memory `project_phase0_foundation_freeze.md`); the remaining work is cron + banner.
- **Why this one over alternatives:** Only Tier 1 compounding pick with HIGH reversibility. Other compounding wins (p0-orchestrator, p0-identity-fn) are MEDIUM rev and slightly larger.
- **Done when:**
  - `/system/freshness.json` updates ≥1 time during 09:00–17:00 ET on a trading day
  - Banner on `/system/` index shows green/yellow/red per property, sourced from freshness.json
  - Simulated 24h-stale playbook output flips its badge to red within 30 min
  - `update_freshness.py` runs via cron without errors for 5 consecutive trading days
- **Files likely touched:** `Z:\public_html\system\update_freshness.py`, `Z:\public_html\system\index.html` (banner injection), Windows Task Scheduler entry (or crontab if on Linux)

---

## Pick 2 — Audit data freshness (read-only diagnostic)

- **Task id:** data-freshness-audit
- **Property:** foundation (INFRA / CROSS-CUTTING)
- **Time horizon:** immediate
- **Estimated effort:** S
- **Tier:** 2 direct (unlock_count=2 + foundation P0)
- **Reversibility:** HIGH (read-only)
- **Leverage justification:** Unlocks `fix-pipeline-0417` (P0) and `fix-missing-days` (P0) directly. User reported newest trades not flowing into DB + Param Surface; if true, this silently invalidates Scorecards, Monthly, Param Surface, and Accountability simultaneously. Read-only diagnostic that should run before any of those properties is trusted again.
- **Why this one over alternatives:** Pairs naturally with Pick 1 — the freshness dashboard surfaces the same staleness signals this audit captures. HIGH-reversibility because there are zero writes.
- **Done when:**
  - Report at `/todos/audit/2026-05-16/freshness-audit.md` lists: last trade_date in main.journal, last datetime_ny in indicator_alerts, last successful Gr8Trade ingest, last pipeline.js run
  - If any source is >24h stale on a trading day, root cause documented + downstream task created
  - Findings feed back into `freshness.json` sources block (consumed by Pick 1's banner)
- **Files likely touched:** report file only (no app code change)

---

## Pick 3 — Identity function (Python + PHP)

- **Task id:** p0-identity-fn
- **Property:** foundation (INFRA / CROSS-CUTTING)
- **Time horizon:** immediate
- **Estimated effort:** M
- **Tier:** 1 compounding
- **Reversibility:** MEDIUM
- **Leverage justification:** Compounding across 4 active properties (scorecards, monthly_reports, accountability_hud, strategy_library). Direct unlock_count=3 (shared-user-dropdown, multi-trader-scorecards, p0-daily-summary-view). Currently `gen_scorecard.py` hardcodes `user_id=2`; `gen_monthly_dashboard.py` is hardcoded to James; accountability-v2 has its own resolution; Strategy Library upload UI is blocked on identity. Replaces 4 hardcoded paths with one helper.
- **Why this one over alternatives:** Highest unlock_count among Tier 1 compounding picks. MEDIUM rev with iterative consumer cutover.
- **Done when:**
  - `/z/setup_daily_reporting/reports/lib/identity.py` exports `resolve_user(account_id) -> (user_id, trader_name)`
  - `/z/public_html/_shared/identity.php` exports `resolve_user($account_id)`
  - gen_scorecard.py, gen_monthly_dashboard.py, accountability-v2 trader-range.php, classify_v2.py all import the helper (zero hardcoded user_id=2 or DEFAULT_ACCOUNT remain in repo)
  - Adding a new trader = INSERT into the mapping table only; no code changes
- **Files likely touched:** `lib/identity.py` (new), `_shared/identity.php` (new), `gen_scorecard.py`, `gen_monthly_dashboard.py`, `accountability-v2/api/trader-range.php`, `classify_v2.py`

---

## Pick 4 — Pipeline orchestrator (run_nightly.py)

- **Task id:** p0-orchestrator
- **Property:** foundation (INFRA / CROSS-CUTTING)
- **Time horizon:** immediate
- **Estimated effort:** M
- **Tier:** 1 compounding
- **Reversibility:** MEDIUM
- **Leverage justification:** Compounding across 5 active properties (scorecards/mgp/monthly_reports/param_surface/playbooks). Retires the entire class of "did this generator run?" silent failures — `mgp-cron-skipped-5-14` is the most recent example. Replaces per-property Task Scheduler entries with one orchestrator that logs to one place and fails loud.
- **Why this one over alternatives:** Pairs with Picks 1 + 2 — the freshness dashboard surfaces what's stale, the orchestrator prevents the staleness in the first place. The cron drift documented 2 days running is a P0 signal that this work is overdue.
- **Done when:**
  - `/z/setup_daily_reporting/run_nightly.py` exists and is invoked by Task Scheduler
  - Log file `/z/logs/nightly-YYYY-MM-DD.log` written per run, includes start+end+exit code per generator
  - A deliberately broken generator (returns empty) causes non-zero exit + email alert within 1 run
  - No per-property "did this run?" diagnostics required for 5 consecutive trading days
- **Files likely touched:** `run_nightly.py` (new), Windows Task Scheduler XML, `/z/logs/` directory

---

## Pick 5 — Param Surface registry pivot

- **Task id:** p0-param-surface-registry
- **Property:** param_surface (OPTIMIZE / POST-TRADE)
- **Time horizon:** immediate
- **Estimated effort:** M
- **Tier:** 1 compounding
- **Reversibility:** MEDIUM
- **Leverage justification:** Compounding across param_surface + strategy_library + scorecards. Unlocks `param-surface-strategy-registry` (P1) and `param-surface-freshness` (P1). Currently Param Surface is a single VWAP grid; new strategies (VWAP Pull-In, Back$ide Scalp, Range-Break) cannot get their own grids without duplicating the surface-generation pipeline.
- **Why this one over alternatives:** Tier 1 with concrete unlock chain. Existing `/param-surface-db/` stays untouched (additive registry); the cutover to `/system/strategies.json` IDs is Picks 6+ (the JS overlay refactor).
- **Done when:**
  - `/param-surface/` index lists ≥3 strategies (VWAP bands, td-vwap, vwap-pull-in) with freshness badge per row
  - Each registry row links to a `/param-surface/<strategy-id>/` grid using the same 3x4 template
  - Registry rows seeded from `/system/strategies.json` so adding a strategy = JSON edit only
  - `main.user_strategies` storage scaffolded but optional (built-ins seeded by code first)
- **Files likely touched:** `/z/public_html/param-surface/index.html` (new), `gen_param_surface.py` (refactor for per-strategy output), `/system/strategies.json` (read-only consumer)

---

## Pick 6 — Refactor _scientific_overlay.js to load from canonical catalog

- **Task id:** refactor-scientific-overlay-js-to-load-from-system
- **Property:** scorecards (GRADE / POST-TRADE)
- **Time horizon:** near_term
- **Estimated effort:** S
- **Tier:** 2 direct (unlock_count=2; supports catalog cutover)
- **Reversibility:** MEDIUM
- **Leverage justification:** First step of the catalog refactor that headlines the supervised queue. Unlocks `refactor-gen-scorecard-py-classify-entry-to-use-re` (P1) and `split-td-vwap-from-vwap-pull-in-in-classifier` (P1). Engine/UI separation per `project_system_contract.md` rule #5.
- **Why this one over alternatives:** Single-file JS change. Loaded fetch contract is the new dependency on `/system/strategies.json` — that file already exists per 5/15 family migration. Lower-risk than the supervised catalog-refactor (which touches 5 surfaces); this is one of those 5.
- **Done when:**
  - `_scientific_overlay.js` fetches `/system/strategies.json` at load + dispatches by strategy id
  - No inline strategy rule constants remain in `_scientific_overlay.js`
  - `/strategies/` page renders identical output before and after
- **Files likely touched:** `Z:\public_html\scorecards\_scientific_overlay.js` (or wherever it lives — verify), possibly `/strategies/` page consumer

---

## Pick 7 — Scorecard intraday equity by time of day

- **Task id:** scorecard-time-of-day-equity
- **Property:** scorecards (GRADE / POST-TRADE)
- **Time horizon:** near_term
- **Estimated effort:** S
- **Tier:** 4 standard (HIGH rev, P1, clean DoD)
- **Reversibility:** HIGH
- **Leverage justification:** Quality improvement to existing scorecard rendering. James's reasoning in the note: "shows Trade 1..101 on x-axis which loses the when-did-I-make-money signal." Single chart axis change in gen_scorecard.py.
- **Why this one over alternatives:** Pure HIGH-reversibility leaf work. Good "after the foundation work, ship one tangible scorecard improvement" pick if cycle time permits.
- **Done when:**
  - Intraday equity curve has x-axis = ET time of day (09:30–16:00)
  - Y-axis stays cumulative P&L
  - Reference scorecard regen matches pre-change Y values at end of day
- **Files likely touched:** `gen_scorecard.py` (axis assignment), `trade_scorecard_*.html` (regen)

---

## Pick 8 — Nightly PM silent-failure detection

- **Task id:** pm-nightly-silent-failure-detection
- **Property:** pm_system (INFRA / CROSS-CUTTING)
- **Time horizon:** near_term
- **Estimated effort:** S
- **Tier:** 4 standard (HIGH rev, P1)
- **Reversibility:** HIGH
- **Leverage justification:** Same theme as Pick 1 (catch silent failures) but scoped to the nightly PM task itself. 5/7 nightly fire produced no output; detection step prevents the next occurrence from going unnoticed.
- **Why this one over alternatives:** Complements Picks 1 + 4 — three layers of silent-failure protection (freshness dashboard, orchestrator, PM self-check). Small, isolated change.
- **Done when:**
  - PM SKILL.md adds a health check at end of run: did /todos/daily/YYYY-MM-DD.md get written? did meta.last_nightly_run advance?
  - If health check fails, alert emitted (email + log)
  - 5 consecutive nightly runs verify the check is a no-op on healthy days
- **Files likely touched:** `C:\Users\jtdom\.claude\scheduled-tasks\nightly-project-manager\SKILL.md`

---

## Pick 9 — MGP daily-stop API wire

- **Task id:** mgp-daily-stop-pipe
- **Property:** daily_stop (LIVE / LIVE COCKPIT)
- **Time horizon:** near_term
- **Estimated effort:** S
- **Tier:** 4 standard (MEDIUM rev, P1)
- **Reversibility:** MEDIUM
- **Leverage justification:** Replaces hardcoded `$500` in MGP Max Risk calc with an API read from `/scorecards/api/user_settings.php` (already shipped per task note). Cleans up a known hardcode in the active MGP pipeline.
- **Why this one over alternatives:** Small, isolated change to gen_mgp.py. Fits naturally if there's bandwidth after the foundation picks.
- **Done when:**
  - MGP Max Risk reads daily_stop from `/scorecards/api/user_settings.php`
  - Fallback to hardcoded value (defensive) if the API read fails
  - Sample MGP regen shows correct Max Risk for James's current stop setting
- **Files likely touched:** `gen_mgp.py` (or the Max Risk computation module)

---

## Lower-priority remaining picks (10-14)

These are EXECUTION_READY and pass all filters but are below the line in tier ranking. Mentioned for completeness:

- **Pick 10** — strategy-vpi-rules (MEDIUM rev, P1, unlocks=3 — codifies VWAP Pull-In). Worth picking up if no foundation appetite. Tier 4 standard.
- **Pick 11** — scorecard-strategy-time-of-day (MEDIUM rev, P1). Companion to Pick 7. Tier 4.
- **Pick 12** — pm-nightly-task (MEDIUM rev, P1). Per cognitive_debt SIMP-1, this should be split into two narrower tasks before execution.
- **Pick 13** — ux-cross-nav (MEDIUM rev, P1). Shared top-nav across 8 properties; per compounding_infrastructure.md, the friction is UX rather than engineering, so this is Tier 4 standard, not Tier 1. Re-qualified from PROBABLY_SHIPPED misfire.
- **Pick 14** — mtf-vp-tradingview-handler (MEDIUM rev, P1). Path B of the MTF VP scanner — webhook endpoint + auth + idempotency. Re-qualified from PROBABLY_SHIPPED misfire.

**Not autonomous-eligible (skipped):** mgp-cron-skipped-5-14 (HIGH rev, P0, owner="James" per drift). Owner casing means this is technically not claude-owned. Route to James, not the Saturday queue.

---

## Summary of autonomous queue

| Pick | id | tier | rev | horizon | est | leverage |
|---|---|---|---|---|---|---|
| 1 | p0-freshness-dashboard | 1 compounding | HIGH | immediate | S | 7 props |
| 2 | data-freshness-audit | 2 direct | HIGH | immediate | S | unlocks 2 P0 |
| 3 | p0-identity-fn | 1 compounding | MEDIUM | immediate | M | 4 props + 3 unlocks |
| 4 | p0-orchestrator | 1 compounding | MEDIUM | immediate | M | 5 props |
| 5 | p0-param-surface-registry | 1 compounding | MEDIUM | immediate | M | 3 props + 2 unlocks |
| 6 | refactor-scientific-overlay-js | 2 direct | MEDIUM | near_term | S | unlocks 2 |
| 7 | scorecard-time-of-day-equity | 4 standard | HIGH | near_term | S | quality |
| 8 | pm-nightly-silent-failure-detection | 4 standard | HIGH | near_term | S | infra |
| 9 | mgp-daily-stop-pipe | 4 standard | MEDIUM | near_term | S | hardcode cleanup |
| 10 | strategy-vpi-rules | 4 standard | MEDIUM | near_term | M | unlocks 3 |
| 11 | scorecard-strategy-time-of-day | 4 standard | MEDIUM | near_term | S | companion to #7 |
| 12 | pm-nightly-task | 4 standard | MEDIUM | near_term | M | split first |
| 13 | ux-cross-nav | 4 standard | MEDIUM | near_term | M | 8-property header |
| 14 | mtf-vp-tradingview-handler | 4 standard | MEDIUM | near_term | M | Path B handler |

**4 Tier-1 compounding picks at the top.** That's not normal — most weeks have 0-1. The audit caught a moment when foundation work is unusually well-positioned. **The Saturday queue is leverage-heavy by design, not by accident.**

A reasonable Saturday cycle: Picks 1+2 in the morning (small, sets up the rest of the work to be verified via the new freshness dashboard), then Pick 3 in the afternoon (the highest-unlock_count pick). Pick 4 or 5 if cycle time remains. **Do not attempt all 9 in one session.**
