---name: cell_agent description: LLM-driven multi-agent framework for automated single-cell analysis. keywords:

• scRNA-seq
• scanpy
• annotation
• autonomous
• bioinformatics

measurable_outcome: Achieves >85% accuracy in cell type annotation compared to manual curation on standard benchmarks. license: MIT metadata: author: Artificial Intelligence Group version: "1.0.0" compatibility:

• system: Python 3.9+

allowed-tools:

• run_shell_command
• read_file

---"

CellAgent

CellAgent is a multi-agent system capable of autonomously handling the entire single-cell RNA-seq (scRNA-seq) analysis pipeline. It simulates a team of biological experts to process data, annotate cells, and perform downstream analysis.

When to Use This Skill

• Automated Annotation: When you have raw scRNA-seq data and need cell type labels without manual curation.
• Complex Workflows: For multi-step analysis (QC -> Clustering -> Annotation -> DE Analysis).
• Data Integration: When merging multiple datasets (e.g., from different batches).

Core Capabilities

1. Planning: Decomposes analysis goals into executable steps.
2. Tool Execution: Generates and runs Python code for Scanpy/Seurat.
3. Self-Correction: detects errors in execution and attempts to fix them.

Workflow

1. Input: User query + scRNA-seq data (H5AD).
2. Planner: The Planning Agent breaks the task into sub-tasks.
3. Executor: The Coding Agent writes scripts to execute the plan.
4. Reviewer: Checks the results and logs outputs.

Example Usage

User: "Process this dataset, filter low-quality cells, and annotate clusters."

Agent Action:

bash
# Assuming a wrapper exists or running the main module from the repo
python3 Skills/Genomics/Single_Cell/CellAgent/repo/main.py --data "./data.h5ad" --goal "annotate"

References

• Mao et al., 2025
• arXiv 2407.09811