Data Analysis
ssgsea-immune-infiltration-analysis
Quantify immune cell infiltration per sample using single-sample GSEA (ssGSEA) with curated immune cell gene signatures. Inputs: expression matrix. Outputs: immune cell score matrix, heatmap, violin comparison plots, correlation with clinical variables.
95100Total Score
Core Capability
94 / 100
Functional Suitability
12 / 12
Reliability
11 / 12
Performance & Context
7 / 8
Agent Usability
15 / 16
Human Usability
7 / 8
Security
12 / 12
Maintainability
12 / 12
Agent-Specific
18 / 20
Medical Task
20 / 20 Passed
96Default ssGSEA with plots
4/4
95GSVA Gaussian without plots
4/4
96Numeric column selectors
4/4
95Named column selectors
4/4
99Reuse-toggle no-plot cleanup
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
PASSResearch Veto✅ PASS — Applicable
| Dimension | Result | Detail |
|---|---|---|
| Scientific Integrity | PASS | Scientific integrity check completed without reportable redline findings. |
| Practice Boundaries | PASS | The skill explicitly excludes clinical decision making and produced only descriptive analytic outputs. |
| Methodological Ground | PASS | The workflow correctly frames ssGSEA/GSVA as relative enrichment and correlation as hypothesis-generating rather than causal inference. |
| Code Usability | PASS | The R pipeline executed successfully across packaged tests, five audit inputs, and the output-directory reuse regression scenario. |
Core Capability94 / 100 — 8 Categories
Functional Suitability
Functional suitability was evaluated against the skill-auditor rubric; no blocking schema-level issue was recorded for this dimension.
12 / 12
100%
Reliability
The stale-plot rerun bug is fixed; only a minor no-plot summary wording ambiguity remains in run_record.txt.
11 / 12
92%
Performance & Context
Canonical plotted runs remain moderately heavy but acceptable for the packaged example data.
7 / 8
88%
Agent Usability
Agent usability was evaluated against the skill-auditor rubric; no blocking schema-level issue was recorded for this dimension.
15 / 16
94%
Human Usability
The skill is easy to follow; the empty plot directory behavior could be documented more explicitly.
7 / 8
88%
Security
Security was evaluated against the skill-auditor rubric; no blocking schema-level issue was recorded for this dimension.
12 / 12
100%
Maintainability
Regression coverage now protects both malformed inputs and output-directory reuse semantics.
12 / 12
100%
Agent-Specific
The skill is now effectively idempotent for reruns; only the no-plot summary wording remains slightly imprecise.
18 / 20
90%
Core Capability Total94 / 100
Medical TaskExecution Average: 96.2 / 100 — Assertions: 20/20 Passed
96
Canonical
Default ssGSEA with plots
4/4 ✓
95
Variant A
GSVA Gaussian without plots
4/4 ✓
96
Edge
Numeric column selectors
4/4 ✓
95
Variant B
Named column selectors
4/4 ✓
99
Stress
Reuse-toggle no-plot cleanup
4/4 ✓
96
Canonical✅ Pass
Default ssGSEA with plots
The full documented output set was produced, including plots, run_record.txt, and output_manifest.txt.
Basic 39/40|Specialized 57/60|Total 96/100
✅A1Required outputs were generated for the audited workflow.
✅A2Input handling and validation behaved as documented.
✅A3No unsupported medical or scientific claim fabrication was detected.
✅A4Execution stayed within the stated Data Analysis skill scope.
Pass rate: 4 / 4
95
Variant A✅ Pass
GSVA Gaussian without plots
The documented GSVA audited baseline completed successfully and the manifest omitted plot entries.
Basic 38/40|Specialized 57/60|Total 95/100
✅A1Required outputs were generated for the audited workflow.
✅A2Input handling and validation behaved as documented.
✅A3No unsupported medical or scientific claim fabrication was detected.
✅A4Execution stayed within the stated Data Analysis skill scope.
Pass rate: 4 / 4
96
Edge✅ Pass
Numeric column selectors
Index-based selector resolution and low-bound runtime settings completed successfully.
Basic 38/40|Specialized 58/60|Total 96/100
✅A1Required outputs were generated for the audited workflow.
✅A2Input handling and validation behaved as documented.
✅A3No unsupported medical or scientific claim fabrication was detected.
✅A4Execution stayed within the stated Data Analysis skill scope.
Pass rate: 4 / 4
95
Variant B✅ Pass
Named column selectors
Named column resolution matched the documented workflow and preserved the expected output schema.
Basic 38/40|Specialized 57/60|Total 95/100
✅A1Required outputs were generated for the audited workflow.
✅A2Input handling and validation behaved as documented.
✅A3No unsupported medical or scientific claim fabrication was detected.
✅A4Execution stayed within the stated Data Analysis skill scope.
Pass rate: 4 / 4
99
Stress✅ Pass
Reuse-toggle no-plot cleanup
After rerunning the same output directory with --make_plots=false, no stale PDFs remained and the manifest had no plot entries.
Basic 40/40|Specialized 59/60|Total 99/100
✅A1Required outputs were generated for the audited workflow.
✅A2Input handling and validation behaved as documented.
✅A3No unsupported medical or scientific claim fabrication was detected.
✅A4Execution stayed within the stated Data Analysis skill scope.
Pass rate: 4 / 4
Medical Task Total96.2 / 100
Key Strengths
- The previous output-directory reuse bug is fixed and now protected by an explicit regression test.
- Generated manifests are now aligned with the documented output inventory, including run_record.txt and output_manifest.txt.
- Repeated canonical runs with the same seed produced identical hashes for key result tables, supporting deterministic behavior.