# Pass 7 — System drift

Date: 2026-05-16

Drift = a task's title/note/path/property no longer matches current vocabulary, file paths, schema, or canonical taxonomy.

## Drift findings (12)

### 1. catalog-refactor-shared-json — **path/title drift** (HIGH leverage)

> Task title: "extract 16-strategy catalog to `/strategies/catalog.json`"
> Task note: "Extract to one `/strategies/catalog.json` fetched by all surfaces."
> **Reality:** The canonical strategy catalog already exists at **`/system/strategies.json`** after the 2026-05-15 family migration. Confirmed in `Z:\public_html\system\` directory listing + memory `reference_canonical_taxonomy.md`.

Other tasks already reference the correct path:
- `refactor-scientific-overlay-js-to-load-from-system` (created 5/12) says "loads from `/system/strategies.json`"
- `refactor-strategies-page-to-render-from-system-str` (created 5/12) says "auto-rendered from `/system/strategies.json`"

Action: update title to "Refactor consumers to load from /system/strategies.json (catalog already canonical)" and note to reflect that the file already exists — remaining work is consumer-side cutover. This is the **biggest single drift finding** because the headline weekend pick recommends it.

### 2. mgp-cron-skipped-5-14 — **owner field convention drift**

> Owner: `"James"` (capitalized)
> Convention: `james` (lowercase), per all other james-owned tasks. The audit found exactly 1 task with `"James"` (this one) + 1 task with `"james"` (mgp-historical-pages-cleanup).

Source: merged via `merge_pending.php` from `/todos/pending_tasks.json` on 2026-05-12. The pending-task merge accepted the capitalized form.

Action: normalize to `james` (lowercase) in the patch.

### 3. Property mis-assignment cluster — **5 tasks**

Tasks have a `property` field that doesn't match what they actually do:

| id | current property | should be |
|---|---|---|
| verify-enrich-on-server | playbooks | scorecards (enrichment runs on gen_scorecard.py path) |
| mgp-multi-tf-bias | scorecards | mgp (task is on MGP Section 1 bias card) |
| mgp-rsi-overextended-rule | playbooks | mgp (task is on MGP overextended classification) |
| strategy-vpi-rules | scorecards | strategy_library or param_surface (registry codification) |
| strategy-backside-rules | scorecards | strategy_library or param_surface |

Pattern: tasks created 2026-04-18/19 grew up before the property taxonomy stabilized. The current 13 active properties (per /system/ hub) are the right home; these need re-homing.

Each mis-assignment artificially inflates the receiving property's pending count and skews the systems.md task_density score for scorecards (currently 27 pending, the highest). Re-homing drops scorecards to ~22, which is closer to its real footprint.

### 4. p0-indicator-alerts-ddl — **path drift**

> Note: "`psql -c '\\d alerts_tradingview.indicator_alerts' > /z/trading/a1p_journal/docs/indicator_alerts_schema.txt`"
> Reality: `/z/trading/a1p_journal/docs/` exists, but consumers of the schema dump (audit-webhook-coverage, wire-webhooks, p0-daily-summary-view) don't read from a fixed path. The output location should match whatever file the downstream tasks expect.

Action: confirm output path with downstream consumer notes before user runs the command.

### 5. p0-daily-summary-view — **schema drift (forward-looking)**

> Note proposes columns: `(user_id, trade_date, pnl, trade_count, win_rate, commissions, ...)`
> **Missing:** `strategy_id` — but a separate task (`add-strategy-id-column-to-indicator-alerts-main-jo`, P0, claude) proposes adding `strategy_id` to `indicator_alerts` + `main.journal` as a stable FK.

If `p0-daily-summary-view` ships before `add-strategy-id-column-...`, the view can't carry strategy attribution, and Strategy Library + Param Surface can't join through it. Either:
- Add `strategy_id` to the view's column list now (forward-looking), or
- Sequence: ship strategy_id column first, THEN ship the view with strategy_id included.

Action: patch.json adds a `depends_on` edge from `p0-daily-summary-view` to `add-strategy-id-column-to-indicator-alerts-main-jo` so they're co-sequenced.

### 6. pm-activity-scanner-deploy — **path display artifact**

> Note: "`Z:\public_html\activity\z_activity_scanner.py`"
> JSON escaping inserted a control char (`\x07` BELL) which makes the path render as `Z:\public_html`(BELL)`ctivity\z_activity_scanner.py` in the todos.json display.

Not a real path drift — display artifact only. The actual file path is correct.

Action: re-write the note in patch.json to use forward slashes or escaped backslashes to avoid the `\a` escape collision.

### 7. acct-v2-bug-risk-bump-ignores-date-range — **evidence freshness drift**

> Bug reported 2026-05-07 with two screenshots; **9 days with no movement.** Bug is current_priority_id per meta but owner is jimbo (correctly — it's a v2 backend bug).

Verify it still reproduces on the current DEV HUD before any work is queued.

Action: smoke-check the repro before treating it as a Saturday pick.

### 8. restore-0416 — **status drift**

> Task: "Restore 4/16 rescued notes into DB"
> Note: "SQL insert after user finishes the 3 steps above."
> The 3 steps (`journal-user-steps`) are 28 days old. If James has run them in the last 4 weeks but not marked them done, this task is silently unblocked. Worth a sanity-check.

### 9. mgp-recap-stability — **task drift** (superseded)

Already addressed in Pass 1 (bucket=SUPERSEDED). Recording here for completeness: `mgp-cron-skipped-5-14` is broader, fresher, and on today_focus[2]; the older task adds no signal.

### 10. risk_hud cluster — **architecture drift**

5 build tasks (`risk-hud-stop-gauge`, `risk-hud-summary-row`, `risk-hud-sort-risk`, `risk-hud-auto-refresh`, `risk-hud-api-wire`) target `/accountability/risk.html` (a mock-data page built 2026-04-21). The new live cockpit at `live-accountability-v2` likely subsumes the use case — `risk-hud-accountability-hud-dedup-review` exists to make this call but hasn't fired yet.

Doing more work into the mock-data version while its replacement is undecided is the canonical operational-confusion smell. Patch.json sets `blocked_by=["risk-hud-accountability-hud-dedup-review"]` on all 5 build tasks.

### 11. multi-trader-scorecards — **priority drift**

Task has `priority=null` but its dependents include `phase-4-trader-selector` (P1) and `phase-4-trader-feedback-loop` (P1). A P1 task should not depend on a no-priority upstream. Bump to P1.

### 12. Old "Chase/Extended" label — **vocabulary drift**

`strategy-align-scorecard-pie` note: "Pie currently shows: Chase/Extended, VWAP Pull-In, Back$ide Scalp, VWAP Pull-In → Trend. 'Chase/Extended' should be a warning class, not a strategy."

This is documented drift — the audit confirms the legacy label needs retirement and gates on `strategy-vpi-rules` + `strategy-backside-rules` being codified first (which themselves are EXECUTION_READY for VWAP Pull-In).

## Compounding drift signal

**Multiple drift findings point at the same root cause: catalog ownership.** 5 of the 12 findings (catalog title drift, p0-daily-summary-view schema drift, strategy property mis-assignments, Chase/Extended label, classifier divergence) all trace back to **the 16-strategy catalog being inlined in 5 places**. That's precisely what `catalog-refactor-shared-json` exists to fix — confirming it as the highest-leverage compounding infrastructure pick.

## Vocabulary drift summary

Built a glossary from the last 30 completed task notes + last 14 daily writeups (top 40 terms). Current canonical vocabulary:

- **gen_mgp.py / gen_mgp_intraday.py** — MGP generators (correct)
- **classify_v2.py** — strategy classifier (note: still referenced as `classify.py` and `_scientific_overlay.js` in 3 places — minor)
- **/system/strategies.json** — canonical catalog (drift: `catalog-refactor-shared-json` says `/strategies/catalog.json`)
- **UNIFIED-FRAMEWORK-v1.md** — 1,165-line doctrine published 5/15
- **family migration** — 2026-05-15 strategies.json migration; tasks created after this date use the correct family vocabulary
- **acct-v2** — `live-accountability-v2`; HUD-off-limits per memory
- **playbook scaffold** — `_modules/momentum_state_engine` + `scanner/`

## Summary

- **12 drift findings.**
- **1 high-leverage drift (catalog-refactor naming)** that affects the headline weekend pick.
- **5 property mis-assignments** — easy fixes, improve task density metrics.
- **1 owner-casing drift** + **1 priority-missing drift** + **1 status-needs-recheck** — housekeeping.
- **The compounding drift signal** points at catalog ownership — supports prioritizing `catalog-refactor-shared-json` even though it's LOW reversibility (supervised queue).
