Variant Pathogenicity Predictor
ReviewAudited by ClawScan on May 10, 2026.
Overview
No evidence of data theft or persistence was found, but the skill should be reviewed because its clinical-genetics claims and documented commands do not match the simple included script.
Review this skill before installing or using it. It does not appear to steal data or persist, but it should not be relied on for clinical decisions: the documentation overstates ACMG/VCF/report capabilities, while the code only calculates a simple score from manual inputs and defaults missing values.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A user could believe the skill performs validated variant interpretation when it does not implement the documented clinical or file-processing functionality.
The user-facing documentation promises VCF/report workflows and clinical-style ACMG/confidence output, but the included script only computes a simple composite score from three optional numeric inputs. This mismatch can cause users or agents to over-trust unsupported medical genetics results.
python scripts/main.py --vcf variants.vcf --output report.json ... Output ... - ACMG guideline interpretation ... - Confidence assessment
Revise the documentation to match the implemented CLI, remove unsupported ACMG/confidence claims, require explicit input scores, and add clear non-clinical-use warnings unless validated clinical logic is implemented.
The tool may output a classification even when the user has not supplied real prediction scores, increasing the chance of misleading interpretation.
The code substitutes default scores when inputs are missing and still emits a pathogenicity classification. In this clinical-genetics context, that can produce authoritative-looking results without variant-specific evidence.
variant = {
"REVEL": args.revel or 0.5,
"CADD": args.cadd or 15,
"PolyPhen": args.polyphen or 0.5
}Fail closed when required scores are missing, clearly label demo/default behavior, and avoid producing clinical-sounding classifications without validated inputs.
It is harder to verify the tool's methodology, authorship, or maintenance status.
The skill has limited provenance information. The included code is small and has no dependencies, so this is a provenance notice rather than evidence of malicious behavior.
Source: unknown Homepage: none
Review the included code and methodology before use, especially because the domain involves medically relevant genetic interpretation.