Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Personal Genomics Analysis
v1.0.1Analyze consumer DNA data from WeGene, 23andMe, AncestryDNA, VCF, BAM, or CRAM files. Generate evidence-based reports covering health risks, pharmacogenomics...
⭐ 1· 103·0 current·0 all-time
byYanzhang@yanzhanglee
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The skill's description and SKILL.md promise a complete pipeline: parsing, SNP lookups (~120 curated SNPs), complex multi-SNP logic (APOE, CYP2C19, MTHFR), report generation, and supplement/clinical recommendations. The file manifest only includes a genotype parser (scripts/parse_genotypes.py) and human-readable references (markdown). There is no programmatic SNP database (described as 'maintained in JSON-compatible format in the scripts directory'), no analysis/interpretation script, and no HTML report generator. Several references (e.g., deep analysis, combining star alleles, report output) are not implemented in the code that is present.
Instruction Scope
SKILL.md explicitly instructs the agent to read user-uploaded genetic files and to elicit sensitive medical history, medications, and family history — which is coherent for genomics analysis but involves highly sensitive personal health/genetic data. The instructions also tell the agent to 'Write a Python script' for analysis and report generation; that implementation is not present, leaving the runtime behavior ambiguous and granting broad discretion. The docs further recommend installing and using samtools/bcftools (including fetching references from remote servers) for CRAM/BAM handling, which implicitly requires network access and additional tools not declared in the manifest.
Install Mechanism
There is no formal install spec (instruction-only), which lowers immediate risk. However references/tool_setup.md contains explicit curl download and compile commands for samtools and bcftools from GitHub releases (legitimate, known release host). That document would cause the environment to download and build external code if followed; this is a moderate-risk operational step (network fetch + compilation) but uses well-known release URLs rather than obfuscated hosts. The repo itself does not automate or declare this install step.
Credentials
The skill declares no required environment variables or credentials (none present in registry metadata). The docs optionally reference setting REF_PATH for CRAM reference fetching, which is reasonable for CRAM handling but is not required. Overall, there are no unexplained credentials or secret requests in the package. Still, the skill will request sensitive personal data (raw genotype files, medical history, medications) as part of normal operation — expected for the stated purpose but privacy-sensitive.
Persistence & Privilege
The skill is not always-enabled and does not request special platform privileges. It does not appear to modify other skills or system-wide settings. No persistent credentials or auto-enablement behavior is present in the manifest.
What to consider before installing
This package documents a full genomics analysis pipeline but only provides a genotype parser and supporting docs — the analysis engine (SNP DB in machine-readable form, multi-SNP logic, and report generator) appears to be missing. Before installing or uploading any personal genomic files: 1) Confirm with the publisher where the analysis/reporting code and the programmatic SNP database live; do not upload sensitive data if the code that performs interpretation is absent or opaque. 2) If you need CRAM/BAM support, note the docs instruct building samtools/bcftools from GitHub (requires network access and compilation); only perform this in an environment you control and trust. 3) Be aware the skill will ask for medical history and medication lists — this is expected for personalized interpretation, but treat it as highly sensitive personal health information. 4) Review the parse_genotypes.py output locally first to ensure data never leaves your machine; ask the maintainer to provide the missing analysis/reporting scripts (or a clear, auditable implementation) before trusting automated recommendations (especially dosing or clinical suggestions). 5) Consider running any untrusted code in an isolated VM/container and consult a qualified clinician for medical decisions; do not treat supplement or diagnostic recommendations as definitive medical advice.Like a lobster shell, security has layers — review code before you run it.
latestvk97fedvr98xe77wkt78n4ceybd8341tj
License
MIT-0
Free to use, modify, and redistribute. No attribution required.