Dossier Generation Skill

Purpose

Generate investment dossiers and IC (Investment Committee) memos from pipeline output. This skill defines the structure, cadence, and content requirements for transforming raw screening results into actionable investment documents.

Preconditions

•Pipeline run has completed successfully (all audit stages OK).
•Governance metadata is present in screening output.
•All scores referenced in the dossier are from the same as_of_date.
•No future data has leaked into the scoring (PIT-verified).

Weekly Cadence

Day	Activity	Owner
Tuesday	Machine processing: full pipeline run, data collection, scoring	Automated
Wednesday	Human scoring: manual review of flagged tickers, override decisions	Analyst
Thursday	IC presentation: ranked portfolio, position sizing, risk overlays	Committee

Dossier Structure

Section 1: Run Summary

code

Run ID:           <deterministic hash>
As-of Date:       <YYYY-MM-DD>
PIT Cutoff:       <as_of_date - 1>
Score Version:    <version>
Schema Version:   <version>
Parameters Hash:  sha256:<hash>

Section 2: Universe Overview

Metric	Value
Total tickers screened	N
Active (passed gates)	N
Excluded (SEV3)	N
Shell companies filtered	N
Below liquidity gate	N
Below market cap gate	N

Section 3: Signal Coverage Dashboard

Report coverage for each signal component. Flag any component below its minimum threshold.

Signal	Coverage	Threshold	Status
Financial scores	X/N (Y%)	80%	OK/DEGRADED
Clinical scores	X/N (Y%)	80%	OK/DEGRADED
Catalyst events	X/N (Y%)	10%	OK/DEGRADED
Market data	X/N (Y%)	0%	OPTIONAL
Momentum signal	X/N (Y%)	0%	OPTIONAL
Valuation signal	X/N (Y%)	0%	OPTIONAL
Smart money (13F)	X/N (Y%)	0%	OPTIONAL
Short interest	X/N (Y%)	0%	OPTIONAL
PoS scores	X/N (Y%)	0%	OPTIONAL

Section 4: Regime Context

Indicator	Value	Regime Signal
VIX level	X.X	Calm/Normal/Elevated/High/Extreme
XBI vs SPY (30d)	X.X%	Outperform/Underperform
Yield curve (10Y-2Y)	X bps	Normal/Inverted
HY credit spread	X bps	Normal/Elevated/Crisis
Classified regime	X	BULL/BEAR/VOLATILITY/etc.

Regime Signal Adjustments Applied

Signal	Multiplier	Rationale
Momentum	X.Xx	Regime-adjusted
Fundamental	X.Xx	Regime-adjusted
Quality	X.Xx	Regime-adjusted
Catalyst	X.Xx	Regime-adjusted
Clinical	X.Xx	Regime-adjusted
Financial	X.Xx	Regime-adjusted

Section 5: Top Holdings (Ranked Portfolio)

For each ticker in the top N (default 20), present:

code

Rank: #X
Ticker: XXXX
Composite Score: XX.XX / 100
Position Size: X.X%

Score Breakdown:
  Clinical:   XX.XX (weight: XX%)  [Phase: X, Indication: X]
  Financial:  XX.XX (weight: XX%)  [Runway: Xmo, Severity: X]
  Catalyst:   XX.XX (weight: XX%)  [Events: X, Proximity: X days]
  PoS:        XX.XX (weight: XX%)  [LOA: X.XXX, Confidence: X.XX]
  Momentum:   XX.XX (weight: XX%)  [Direction: X]
  Valuation:  XX.XX (weight: XX%)  [Peer rank: X]
  Short Int:  XX.XX (weight: XX%)  [SI: X%, DTC: X, Signal: X]

Enhancement Flags:
  - E1 Regime Gating:    [applied/not applied]
  - E2 Existential Flaw: [applied/not applied] [cap: XX]
  - E3 Confidence Gate:  [components gated: X]
  - E4 Score Ceiling:    [ceiling: XX, reason: X]
  - E5 Asymmetric:       [transform applied: X]
  - E6 Contradictions:   [conflicts: X]

Risk Flags:
  - Dilution Risk: [NO_RISK/LOW/MEDIUM/HIGH] (score: X.XX)
  - Liquidity Risk: [flags]
  - Crowding Risk: [LOW/MEDIUM/HIGH]
  - Staleness: [PASS/WARN/SOFT_GATE]

Smart Money:
  - Managers holding: X/33 elite
  - Coordinated activity: [ADD/EXIT/NONE]
  - Crowding: [YES/NO] (threshold: 6+)

Section 6: Exclusions Report

List all excluded tickers with reasons:

Ticker	Exclusion Reason	Gate
XXXX	Runway < 6 months	SEV3 (financial)
XXXX	ADV < $500K	Liquidity hard gate
XXXX	Shell company	Universe filter
XXXX	Market cap < $50M	Size filter

Section 7: Score Distribution

Report distribution statistics for each score component across the rankable universe:

Component	Mean	Median	Std Dev	Min	Max	Skew
Composite	X.X	X.X	X.X	X.X	X.X	X.X
Clinical	X.X	X.X	X.X	X.X	X.X	X.X
Financial	X.X	X.X	X.X	X.X	X.X	X.X
Catalyst	X.X	X.X	X.X	X.X	X.X	X.X

Section 8: Week-over-Week Changes

Compare current run to previous week's run:

Change Type	Count	Details
New to top 20	N	[tickers]
Dropped from top 20	N	[tickers]
Rank change > 10	N	[tickers with direction]
New exclusions	N	[tickers with reasons]
Re-included	N	[tickers]
New catalyst events	N	[tickers with event types]

Section 9: Position Sizing Summary

Metric	Value
Total positions	N (max: 60)
Target weight sum	100%
Max single position	X.X% (limit: 10%)
Min single position	X.X% (limit: 0.5%)
HHI (concentration)	X.XXXX

Composite Weight Sets

Document which weight set was used and why:

V3 Enhanced (all signals available)

Component	Weight
Clinical	26%
Financial	24%
Catalyst	16%
PoS	14%
Momentum	9%
Short Interest	6%
Valuation	5%

V3 Default (no enhancement data)

Component	Weight
Clinical	40%
Financial	35%
Catalyst	25%

V3 Partial (some enhancements)

Component	Weight
Clinical	33%
Financial	28%
Catalyst	18%
Momentum	9%
Short Interest	7%
Valuation	5%

Baker-Style Fundamental-Concentrated

Component	Weight	Rationale
Clinical	35%	Core thesis (biology quality)
PoS	18%	Core thesis (probability)
Financial	22%	Survivability
Valuation	15%	Mispricing
Catalyst	7%	Timing only
Momentum	2%	Overlay
Short Interest	1%	Risk context

Enhancement Flags Reference (E1-E6)

E1: Hard Regime Gating

Regime	Momentum Cap	Catalyst Adj	Financial Penalty
BEAR	30% of deviation	-	1.25x
BULL	100% (full)	1.15x boost	0.85x
NEUTRAL	No gating	-	-

E2: Existential Flaw Escalation

•Runway < 9 months AND early-stage (phase_1, phase_2): cap score at 65
•Alternative: apply 20% penalty

E3: Confidence-Weighted Aggregation

Per-component confidence multipliers. Components below 0.40 confidence are soft-gated (weight reduced, not zeroed).

E4: Dynamic Score Ceilings

Condition	Ceiling
Phase 3 stage	<= 85
No catalyst events	<= 60
Commercial stage	<= 90

E5: Convex Downside, Concave Upside

•Negative deviations from neutral are amplified (convex loss function)
•Positive deviations are dampened (concave gain function)
•Reflects loss aversion appropriate for biotech risk

E6: Contradiction Detector

Conflict Pair	Resolution
Strong momentum + low liquidity	Cap momentum contribution
Cheap valuation + financing pressure	Flag, reduce conviction

Smart Money Section Detail

Manager Tier Weights

Tier	Weight	Description
Elite Core	1.5x	Top-tier biotech specialists
Conditional	1.0x	Other tracked managers

Position Change Scoring

Action	Share Change	Score
NEW	0 -> positive	+10
ADD	> +10%	+5
HOLD	+/- 10%	+2
TRIM	> -10%	-3
EXIT	positive -> 0	-8

Coordinated Activity

Signal	Threshold	Meaning
Coordinated buying	>= 3 managers adding	Bullish conviction
Fresh conviction	>= 2 managers new positions	Strong bullish
Crowding	>= 6 managers holding	Potential crowding risk

Expected Return Model

For translating composite scores into expected returns:

code

Score -> Rank -> Percentile (Blom plotting position)
Percentile -> Z-score (Acklam inverse normal)
Z-score -> Expected excess return = z * lambda * (12 / holding_period_months)

•DEFAULT_LAMBDA_ANNUAL: 0.08 (8% per 1-sigma per year)
•Model ID: zscore_linear_lambda
•Model Version: 1.0.0

Output Format Requirements

•All monetary values in USD with commas (e.g., $1,234,567)
•All scores to 2 decimal places
•All percentages to 1 decimal place
•All dates in ISO 8601
•Governance block attached to every output document
•Content hash (SHA256) of the dossier included in audit log

Source Files

Component	File
Pipeline Orchestrator	`run_screen.py`
Composite Scoring	`module_5_composite_v3.py`
Scoring Types	`module_5_scoring_v3.py`
Diagnostics	`module_5_diagnostics_v3.py`
Regime Engine	`regime_engine.py`
Smart Money	`manager_momentum_v1.py`
Expected Returns	`common/score_to_er.py`
Audit Log	`governance/audit_log.py`
Position Sizing	`module_5_composite_with_defensive.py`