☑ Tasks
Data: 8.23.26 · CRITICAL + HIGH open tasks
Open Tasks
CRITICAL
[INVESTIGATE] BF Type 1 combo scores measure held/sold binary not actual returns — all combo rankings are potentially co…
CRITICAL
[INVESTIGATE] FORMAL recs UNDERPERFORMING control group at p<0.05 — confirmed 3 consecutive independent pipeline runs
HIGH
[INVESTIGATE] Session-start digest block missing from hook injection
HIGH
[ASSESS] Reconcile checkpoint-date discrepancy on 2026-07-12 pipeline kill
HIGH
[BUILD] soldtrack.html long-term size management — prevent hitting Cloudflare 25 MiB hard limit
HIGH
[INVESTIGATE] Session digest token audit — index all session startup injection items and their token cost
HIGH
[GOVERNANCE] Add Skill() built-in command exclusion rule to ~/.claude/CLAUDE.md
HIGH
[GOVERNANCE] Auto-snapshot before every code edit — `.bak` rollback discipline must be machine-enforced, not remembered
Latest Changelog
## 2026-08-23 — Full pipeline run on new CSVs (Holdings 72 / NR 16 / SM 71); Python 3.14 env was missing numpy/pandas/scipy — installed; 4 non-fatal step failures found
**Session:** Confirmed finviz + ratio pulls both completed same-day (`finviz_pull_20260823_0151.csv`, `ratio_pull_20260823_0445.csv`). User dropped 3 new source CSVs (`Stock score (holdings) (72).csv`, `New Research Tab 3.20.24 (16).csv`, `Secondary Metrics (71).csv`) into `data/raw/` and ran the full pipeline.
- **First run failed at Phase 1 Ingest** — `ModuleNotFoundError: No module named 'numpy'`. Root cause: the `py` launcher resolves to a Python 3.14 interpreter (`AppData\Local\Programs\Python\Python314`) that had **zero packages installed** — not a data/pipeline bug, a bare interpreter. Installed numpy, pandas, scipy on that interpreter; re-ran clean (`py run_pipeline.py --yes` — note `--resume` is NOT a valid flag on this pipeline version, memory `feedback_long_running_pulls.md` was stale on that point).
- **Second run completed successfully (exit code 0)**, full 12:20→12:40 run: ingest, direction locks, cohort stats, gate/direction advisors, MPS scoring, BF combos (L2-only + L1+L2 combined), dashboards, mobile deploy (22 files → https://claudemobiledashboards.pages.dev/) all clean.
- **4 non-fatal step failures surfaced (pipeline correctly continued past each):**
1. **Phase 3 L1 Formula Discovery** (`mde_05_l1_formula_discovery.py`) — real bug: `ZeroDivisionError` in the `safe_ab` lambda (`a / (b+1)`) when `b == -1`. Filed as task below.
2. Sold-Ticker Price Fetcher — `yfinance` not installed on the Python 3.14 env
3. Control Lab price fetch — `requests`/`beautifulsoup4` not installed
4. Holdings performance fetch — `yfinance` not installed
(2–4 are the same root cause as the numpy gap — this interpreter never had the full package set. User has not yet decided whether to install them.)
System Health
0 errors · 10 warnings
All warnings: hardcoded dist weights — expected