Data Analysis

nomogram-construction

Use when constructing a prognosis nomogram from survival-related clinical predictors, exporting the nomogram bundle and C-index table, and optionally rendering the final nomogram PDF. NOT for: univariate/multivariable Cox feature screening, calibration curves, ROC analysis, decision-curve analysis, or non-survival outcomes.

96100Total 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

11 / 12

Maintainability

12 / 12

Agent-Specific

20 / 20

Medical Task

20 / 20 Passed

97Build nomogram from age, stage, risk — 3 predictors, 1/2/3 year horizons

5/5

94Build nomogram with 4 predictors including categorical variables

4/4

93Clinical CSV with only 15 complete samples and 5 events — below minimum threshold

3/3

96Render nomogram PDF from existing .qs bundle with custom dimensions and font size

4/4

94Build nomogram with 8 predictors, 1/3/5 year horizons, and 120-second timeout

4/4

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

Dimension	Result	Detail
Scientific Integrity	PASS	No fabricated DOI, PMID, C-index values, or survival statistics; all values computed from provided clinical data via rms::cph.
Practice Boundaries	PASS	Explicitly not for clinical diagnosis; skill is scoped to nomogram construction from pre-selected predictors only.
Methodological Ground	PASS	rms::cph is the canonical Cox model implementation for nomogram construction in R; C-index computation is methodologically correct; minimum sample/event requirements prevent invalid model fitting.
Code Usability	PASS	All 6 R modules syntactically valid; withCallingHandlers/tryCatch pattern correct; qs serialization for portable bundle storage.

Core Capability98 / 100 — 8 Categories

Functional Suitability

Full coverage of nomogram construction, C-index computation, .qs bundle export, and PDF rendering. Two-mode design cleanly separates expensive fitting from plot iteration.

12 / 12

100%

Reliability

Validation for file existence, CSV format, min 20 complete samples, min 10 events; 9 distinct SKILL_* codes covering all failure modes including SKILL_ANALYSIS_ERROR for cph() failures.

12 / 12

100%

Performance & Context

When-to-read table; references deferred; SKILL.md ~340 lines well within acceptable range; two-mode design enables lightweight plot rerenders.

8 / 8

100%

Agent Usability

Mode-based execution is clear; prerequisite section explicit; feedback design note: no append-only run_record/output_manifest like estimate and GSVA sibling skills — only session_info.txt.

15 / 16

94%

Human Usability

Specific trigger phrases (prognosis nomogram, C-index, pre-selected prognostic variables); strict CSV validation correct per Scene Override.

8 / 8

100%

Security

No hardcoded secrets; CSV and column validation present; missing explicit privacy note for clinical CSV input.

11 / 12

92%

Maintainability

6-module R architecture; smoke test with included data and automated script; --overwrite flag enables clean test reruns.

12 / 12

100%

Agent-Specific

Trigger precision strong with pre-selected predictors qualifier; .qs bundle enables clean downstream composition with calibration-curve and plot-only skills; --overwrite guard for idempotency.

20 / 20

100%

Core Capability Total98 / 100

Medical TaskExecution Average: 94.8 / 100 — Assertions: 20/20 Passed

Canonical

Build nomogram from age, stage, risk — 3 predictors, 1/2/3 year horizons

5/5 ✓

Variant A

Build nomogram with 4 predictors including categorical variables

4/4 ✓

Edge

Clinical CSV with only 15 complete samples and 5 events — below minimum threshold

3/3 ✓

Variant B

Render nomogram PDF from existing .qs bundle with custom dimensions and font size

4/4 ✓

Stress

Build nomogram with 8 predictors, 1/3/5 year horizons, and 120-second timeout

4/4 ✓

Canonical✅ Pass

Build nomogram from age, stage, risk — 3 predictors, 1/2/3 year horizons

Full build: validate CSV -> complete-case filter -> cph() -> nomogram() -> C-index -> .qs + .rds + .xlsx. All 4 expected output files generated.

Basic 40/40|Specialized 57/60|Total 97/100

✅A1Nomogram bundle (data/Nomogram_list.qs) saved with all required objects

✅A2C-index table (table/nomogram_c_index.xlsx) generated with metric and value columns

✅A3set.seed() applied before Cox model fitting

✅A4session_info.txt written with build parameters for reproducibility

✅A5No medical diagnosis or clinical recommendation made

Pass rate: 5 / 5

Variant A✅ Pass

Build nomogram with 4 predictors including categorical variables

4 predictors including categorical (gender, treatment) converted to factors automatically. C-index computed for 4-predictor model.

Basic 39/40|Specialized 55/60|Total 94/100

✅A1Character predictors automatically converted to factors for Cox model fitting

✅A2C-index computed correctly for 4-predictor model

✅A3At least 3 features requirement satisfied by 4 predictors

✅A4SKILL_ANALYSIS_ERROR would surface if cph() fitting fails

Pass rate: 4 / 4

Edge✅ Pass

Clinical CSV with only 15 complete samples and 5 events — below minimum threshold

SKILL_INSUFFICIENT_DATA raised: requires >= 20 complete samples and >= 10 events. No partial model saved. Error message identifies specific requirements.

Basic 38/40|Specialized 55/60|Total 93/100

✅A1SKILL_INSUFFICIENT_DATA raised when fewer than 20 complete samples are present

✅A2Error message identifies specific sample and event count requirements

✅A3No partial nomogram model is saved when minimum requirements are not met

Pass rate: 3 / 3

Variant B✅ Pass

Render nomogram PDF from existing .qs bundle with custom dimensions and font size

Plot mode: load .qs bundle -> validate required objects -> render PDF with custom width=12 and font_size=10.

Basic 40/40|Specialized 56/60|Total 96/100

✅A1Plot mode reads from .qs bundle without re-running Cox analysis

✅A2Custom plot_width and font_size parameters applied to PDF rendering

✅A3session_info.txt written in plot directory with plot parameters

✅A4SKILL_FILE_NOT_FOUND raised if .qs bundle is missing

Pass rate: 4 / 4

Stress✅ Pass

Build nomogram with 8 predictors, 1/3/5 year horizons, and 120-second timeout

8 predictors accepted; custom years 1,3,5 parsed; timeout enforced at 120s via apply_time_limit.

Basic 39/40|Specialized 55/60|Total 94/100

✅A18 predictors accepted (minimum 3 required)

✅A2Custom years 1,3,5 correctly parsed and applied to nomogram survival functions

✅A3SKILL_TIMEOUT raised if 120-second limit is exceeded

✅A4No fabricated C-index values — all statistics computed from actual data

Pass rate: 4 / 4

Medical Task Total94.8 / 100

Key Strengths

rms::cph is the methodologically correct foundation for Cox-based nomograms in R; C-index computation is standard and well-validated.
Two-mode separation (build vs plot) with .qs bundle enables efficient iteration on visualization without re-running expensive Cox fitting — excellent design for clinical researchers.
Comprehensive SKILL_* error table with 9 distinct codes including SKILL_ANALYSIS_ERROR covering cph() fitting failures that other skills may silently swallow.
--overwrite guard prevents accidental output directory corruption during repeated test runs.