| # | Priority | Patch | Why |
|---|---|---|---|
| 1 | High | Rule #13 graduated cap Band-A hard cap 60 / Band-B soft cap 75 |
Binary cap is inert for tickers with SBC 12–15%. PLTR is the test case. |
| 2 | High | Accrual ratio decomposition Split operating (ΔWC) vs non-operating (impairments) |
TDOC goodwill writedowns fire the accrual flag incorrectly — non-operating contamination. |
| 3 | Medium | IMPAIRMENT_EVENT tag promotion From informational to halt-taxonomy category |
Currently orphan; needs numeric trigger & single source of truth. |
| 4 | Medium | healthtech adapter → Production | BETA until 4 clean runs logged (DOCS, HIMS, EVH, AMWL). |
| 5 | Medium | Reference portfolio seed 10-ticker basket for calibration |
Score median = 50 needs an anchored, not theoretical, distribution. |
| 6 | Low | Halt taxonomy cross-check METHODOLOGY.md ↔ SKILL.md |
Two files list halts; collapse to single source of truth. |
Agreed stack is Bloomberg + AlphaSense + SEC EDGAR. Question is which subscription is first, and whether Bloomberg is Band A (quant) only or also Band B (qualitative).
Pineal strategy says 8–12 concentrated names. FPR calibration works better with a wider reference portfolio. Question is whether to run the module monthly across a wider watchlist or strictly on active names.
Private mode removes sell-side consensus inputs, loosens N/A tolerance, uses parameterised assumptions. Useful for the PE-style theses Pineal runs. Currently out of v1 scope.
The TDOC run halted at 52% coverage — correct behaviour. But on Tier 2 data the threshold becomes easier to hit, and the module will produce more output. Should the coverage gate stay at 60%, or scale with data tier (e.g., 50% on Tier 4, 75% on Tier 1)? Over-discipline kills throughput; under-discipline kills credibility.
Score bands are A (85+), B (70–84), C (55–69), D (40–54), F (<40). Median is calibrated to 50. Question: is the pass-the-filter threshold a B, or does Pineal want to engage with Cs where the catalyst is strong enough? The answer shapes how aggressive the EQ cap should be — and whether Band-B graduated cap is the right design.
Rule #5 makes the alpha signal mandatory — our 3-year forecast vs sell-side consensus. That's the variant view. Question: does it feed Section 5 of the IC paper (returns/capital) or Section 6 (valuation), and who owns adjudication when our forecast and consensus materially disagree? Both are defensible; the answer shapes the module's downstream integration.