CRISPR_Design_Agent

---name: crispr-design-agent

SKILL.md

GitHub
<!-- # COPYRIGHT NOTICE # This file is part of the "Universal Biomedical Skills" project. # Copyright (c) 2026 MD BABU MIA, PhD <md.babu.mia@mssm.edu> # All Rights Reserved. # # This code is proprietary and confidential. # Unauthorized copying of this file, via any medium is strictly prohibited. # # Provenance: Authenticated by MD BABU MIA -->

name: crispr-guide-design description: Guide foundry keywords: - CRISPR - sgRNA - Doench - off-target - oligos measurable_outcome: Return the requested number of guides (default ≥4) with efficiency + specificity scores, coordinates, and cloning oligos within 10 minutes per gene. license: MIT metadata: author: CRISPR-GPT Team version: "1.0.0" compatibility: - system: Python 3.10+ allowed-tools: - run_shell_command - read_file

CRISPR Design Agent

Automate sgRNA selection, scoring, off-target evaluation, and oligo generation for CRISPR experiments using the documented workflow.

When to Use

  • Designing CRISPR knockout/knock-in experiments that need validated guides.
  • Locating all PAM-compatible target sites in a gene or locus.
  • Filtering guides by efficiency/off-target metrics before cloning.

Core Capabilities

  1. Target discovery: Scan sequences for PAM motifs (e.g., NGG).
  2. Efficiency scoring: Evaluate GC content, homopolymers, Doench/DeepCRISPR/CFD scores.
  3. Filtering & ranking: Remove risky guides (SNP overlap, off-target hits) and output the best candidates.

Workflow

  1. Resolve gene symbol + organism to canonical transcript coordinates and target region.
  2. Enumerate PAM-compatible sites; extract spacers for the chosen Cas variant.
  3. Score guides (efficiency + specificity) and compute GC metrics.
  4. Run off-target search (≤3 mismatches) to flag problematic loci.
  5. Filter/rank guides, generate cloning oligos/primers, and emit JSON/CSV outputs with coordinates.

Example Usage

python3 Skills/Genomics/CRISPR_Design_Agent/crispr_designer.py \
    --sequence "ATGGAGGAGCCGCAGTCAGATCCTAGCGTCGAGCCCCCTCTGAGTCAGGAAACATTTTCAGACCTATGGAAACTGTGAGTGGATCCATTGGAAGGGC" \
    --output guides.json

Guardrails

  • Always state genome build and Cas variant assumptions.
  • Avoid guides overlapping common SNPs when avoid_variants is true.
  • Flag high off-target density near coding regions for manual review.

References

  • See README.md and prompt.md for detailed schema plus supporting literature.
<!-- AUTHOR_SIGNATURE: 9a7f3c2e-MD-BABU-MIA-2026-MSSM-SECURE -->