bio-atac-seq-atac-peak-calling

Pass

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads the ENCODE blacklist, which is a standard reference dataset for genomic analysis, from the Boyle Lab's GitHub repository. This is a common and safe practice in bioinformatics workflows to filter out artifact-prone regions.
  • Evidence: wget https://github.com/Boyle-Lab/Blacklist/raw/master/lists/hg38-blacklist.v2.bed.gz in SKILL.md and usage-guide.md.
  • [COMMAND_EXECUTION]: The skill executes standard bioinformatics command-line tools including macs2, macs3, samtools, idr, and bedtools to perform genomic data processing. These operations are aligned with the skill's primary purpose.
  • Evidence: Invocations of macs2 callpeak, samtools view, idr --samples, and bedtools intersect within examples/call_atac_peaks.sh.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests external genomic data files (BAM and BED formats), which represents a surface for processing untrusted data.
  • Ingestion points: Input variables in examples/call_atac_peaks.sh such as REP1, REP2, and BLACKLIST.
  • Boundary markers: None present, as these are binary or structured data formats not typically containing natural language instructions.
  • Capability inventory: File system writes (mkdir, sort >) and execution of various bioinformatics CLI tools.
  • Sanitization: The script uses set -euo pipefail for shell safety but does not perform content validation on the genomic files themselves, which is standard for this domain.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 8, 2026, 12:03 PM