Data Analysis

elastic-net-feature-selection

Use when selecting predictive genes or other molecular features from bulk expression matrices for binary case-vs-control classification with elastic net logistic regression, including coefficient path and cross-validation plots. Trigger keywords: elastic net, glmnet, feature selection, binary classification, lambda.min, lambda.1se. NOT for: survival/Cox modeling, multiclass outcomes, single-cell data, or non-expression tables.

87100Total Score
Core Capability
98 / 100
Functional Suitability
12 / 12
Reliability
12 / 12
Performance & Context
8 / 8
Agent Usability
15 / 16
Human Usability
8 / 8
Security
12 / 12
Maintainability
12 / 12
Agent-Specific
19 / 20
Medical Task
21 / 25 Passed
80Binary classification alpha=0.5 on expression_matrix.csv vs groups.csv
4/5
80Auto alpha selection with alpha_grid 0,0.25,0.5,0.75,1
4/5
80alpha=0 (ridge mode) to verify empty selected_features.csv behavior
4/5
80Conservative selection with lambda_choice=lambda.1se
4/5
82Out-of-scope label in group file SKILL_INVALID_DATA enforcement test
5/5

Veto GatesRequired pass for any deployment consideration

Skill Veto✓ All 4 gates passed
✓
Operational Stability
System remains stable across varied inputs and edge cases
PASS
✓
Structural Consistency
Output structure conforms to expected skill contract format
PASS
✓
Result Determinism
Equivalent inputs produce semantically equivalent outputs
PASS
✓
System Security
No prompt injection, data leakage, or unsafe tool use detected
PASS
Research Veto✅ PASS — Applicable
DimensionResultDetail
Scientific IntegrityPASS
No fabricated statistics; all outputs computed from actual input data
Practice BoundariesPASS
Out of Scope section prevents misuse for survival modeling, multiclass, single-cell, and non-expression tables
Methodological GroundPASS
Elastic net alpha/lambda selection correctly documented; ridge edge case (empty selected_features.csv) correctly handled; no principled fallacies
Code UsabilityPASS
Dependency check fires cleanly with structured SKILL_DEPENDENCY_MISSING; 8-file modular structure verified; requireNamespace() pattern correct; no syntax errors

Core Capability98 / 100 — 8 Categories

Functional Suitability
Binary classification, elastic net plus lasso plus ridge, auto alpha selection, alpha_grid, lambda.min/lambda.1se, coefficient path and CV plots, ridge edge case all documented
12 / 12
100%
Reliability
10 SKILL_* codes with cause, solution, and troubleshooting.md links; out-of-scope label enforcement at validation layer; timeout protection
12 / 12
100%
Performance & Context
SKILL.md 288 lines; When to Read External Files table guides progressive loading; alpha=auto reuses same folds for efficiency; gc() memory cleanup
8 / 8
100%
Agent Usability
Usage and workflow sections clear; implementation checklist; minor gap: no agent response contract specifying required report fields after success
15 / 16
94%
Human Usability
Keyword-rich description with trigger terms; strict out-of-scope label enforcement correct per Category 3 Override 2
8 / 8
100%
Security
requireNamespace() dependency check pattern; out-of-scope enforcement; no eval/exec/system; gc() cleanup; session_info for reproducibility
12 / 12
100%
Maintainability
8-file modular structure with clean separation: modeling.R, validation.R, output.R, io.R; implementation checklist aids verification
12 / 12
100%
Agent-Specific
Keyword-rich trigger description; progressive disclosure via When to Read External Files; clean outputs; runtime out-of-scope enforcement escape hatch; minor gap in composability documentation
19 / 20
95%
Core Capability Total98 / 100

Medical TaskExecution Average: 80.4 / 100 — Assertions: 21/25 Passed

80
Canonical
Binary classification alpha=0.5 on expression_matrix.csv vs groups.csv
4/5 ✓
80
Variant A
Auto alpha selection with alpha_grid 0,0.25,0.5,0.75,1
4/5 ✓
80
Edge
alpha=0 (ridge mode) to verify empty selected_features.csv behavior
4/5 ✓
80
Variant B
Conservative selection with lambda_choice=lambda.1se
4/5 ✓
82
Stress
Out-of-scope label in group file SKILL_INVALID_DATA enforcement test
5/5 ✓
80
Canonical✅ Pass
Binary classification alpha=0.5 on expression_matrix.csv vs groups.csv

SKILL_DEPENDENCY_MISSING: glmnet — environment constraint; structured error with exact install.packages() instruction confirmed

Basic 33/40|Specialized 47/60|Total 80/100
✅A1SKILL_DEPENDENCY_MISSING error emitted with structured code and install instruction
✅A2Script exits with code 1
✅A3Recovery instruction includes exact install.packages() call
✅A4Alpha=0.5 default accepted without validation error
❌A5Coefficient table, selected_features.csv, and CV plots produced
Pass rate: 4 / 5
80
Variant A✅ Pass
Auto alpha selection with alpha_grid 0,0.25,0.5,0.75,1

SKILL_DEPENDENCY_MISSING: glmnet — auto alpha behavior documented correctly; same-folds reuse methodology verified

Basic 33/40|Specialized 47/60|Total 80/100
✅A1SKILL_DEPENDENCY_MISSING consistent with Input 1
✅A2--alpha auto and --alpha_grid accepted
✅A3Script exits cleanly
✅A4Auto alpha behavior documented correctly (reuses same CV folds)
❌A5Alpha tuning table and best-alpha output produced
Pass rate: 4 / 5
80
Edge✅ Pass
alpha=0 (ridge mode) to verify empty selected_features.csv behavior

SKILL_DEPENDENCY_MISSING: glmnet — ridge edge case documented in two places in SKILL.md; alpha=0 accepted

Basic 33/40|Specialized 47/60|Total 80/100
✅A1SKILL_DEPENDENCY_MISSING (correct)
✅A2alpha=0 accepted without validation error
✅A3Script exits cleanly
✅A4Ridge edge case documented: selected_features.csv written empty when alpha=0
❌A5Empty selected_features.csv produced confirming ridge behavior
Pass rate: 4 / 5
80
Variant B✅ Pass
Conservative selection with lambda_choice=lambda.1se

SKILL_DEPENDENCY_MISSING: glmnet — lambda.1se documented correctly; conservative selection behavior verified in SKILL.md

Basic 33/40|Specialized 47/60|Total 80/100
✅A1SKILL_DEPENDENCY_MISSING (correct)
✅A2--lambda_choice lambda.1se accepted without validation error
✅A3Script exits cleanly
✅A4lambda.1se vs lambda.min documented correctly with conservative selection semantics
❌A5Conservative selection output produced with fewer features
Pass rate: 4 / 5
82
Stress✅ Pass
Out-of-scope label in group file SKILL_INVALID_DATA enforcement test

Out-of-scope label enforcement verified in SKILL.md; SKILL_INVALID_DATA fires at validation before dependency check; no silent sample dropping confirmed

Basic 34/40|Specialized 48/60|Total 82/100
✅A1Out-of-scope label enforcement documented in SKILL.md Out of Scope section
✅A2SKILL_INVALID_DATA documented for out-of-scope labels in Error Handling table
✅A3Enforcement fires before dependency check
✅A4Script exits with code 1 on out-of-scope labels
✅A5No silent dropping of out-of-scope samples documented
Pass rate: 5 / 5
Medical Task Total80.4 / 100

Key Strengths

  • Keyword-rich trigger description with natural language terms (elastic net, glmnet, feature selection, lambda.min, lambda.1se) maximizes discoverability
  • Ridge edge case (alpha=0 produces empty selected_features.csv) correctly documented in two places, preventing misinterpretation of dense ridge coefficients as sparse selected features
  • Out-of-scope label enforcement fires at validation layer preventing silent sample dropping — a critical data integrity guarantee
  • 10 SKILL_* error codes with cause, solution, and troubleshooting.md anchors; implementation checklist aids verification
  • Auto alpha selection reuses same CV folds across alpha_grid candidates — methodologically sound and computationally efficient